transfer_learning.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "display_name": "Python 3",
      "language": "python",
      "name": "python3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.8.5"
    },
    "colab": {
      "name": "transfer_learning.ipynb",
      "provenance": [],
      "collapsed_sections": []
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "p_StsEH0HdKp",
        "outputId": "965173d2-1053-4c83-b876-44477deba134"
      },
      "source": [
        "%tensorflow_version 2.x\n",
        "import tensorflow as tf\n",
        "device_name = tf.test.gpu_device_name()\n",
        "if device_name != '/device:GPU:0':\n",
        "  raise SystemError('GPU device not found')\n",
        "print('Found GPU at: {}'.format(device_name))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Found GPU at: /device:GPU:0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AOV-SR4qIPB6",
        "outputId": "cf4fafab-7434-4a1e-efe7-ad8e59e13f87"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/gdrive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Mounted at /content/gdrive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "kra4LMNXHYlr"
      },
      "source": [
        "## Imports and global parameters"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "1NcXfSt3Hva0",
        "outputId": "df4c5ea2-054e-4b29-e78e-5d6e9ed0129f"
      },
      "source": [
        "!pip install noisereduce"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting noisereduce\n",
            "  Downloading https://files.pythonhosted.org/packages/04/d0/4e50cac3daaa1522a3730ec22750ca86f6c221a480e65c8d6b0ab18a21ed/noisereduce-1.1.0.tar.gz\n",
            "Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from noisereduce) (1.4.1)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.7/dist-packages (from noisereduce) (3.2.2)\n",
            "Requirement already satisfied: librosa in /usr/local/lib/python3.7/dist-packages (from noisereduce) (0.8.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from noisereduce) (1.19.5)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from noisereduce) (4.41.1)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->noisereduce) (2.4.7)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->noisereduce) (1.3.1)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.7/dist-packages (from matplotlib->noisereduce) (2.8.1)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.7/dist-packages (from matplotlib->noisereduce) (0.10.0)\n",
            "Requirement already satisfied: scikit-learn!=0.19.0,>=0.14.0 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (0.22.2.post1)\n",
            "Requirement already satisfied: audioread>=2.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (2.1.9)\n",
            "Requirement already satisfied: decorator>=3.0.0 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (4.4.2)\n",
            "Requirement already satisfied: pooch>=1.0 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (1.3.0)\n",
            "Requirement already satisfied: numba>=0.43.0 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (0.51.2)\n",
            "Requirement already satisfied: resampy>=0.2.2 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (0.2.2)\n",
            "Requirement already satisfied: joblib>=0.14 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (1.0.1)\n",
            "Requirement already satisfied: soundfile>=0.9.0 in /usr/local/lib/python3.7/dist-packages (from librosa->noisereduce) (0.10.3.post1)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.1->matplotlib->noisereduce) (1.15.0)\n",
            "Requirement already satisfied: appdirs in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa->noisereduce) (1.4.4)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa->noisereduce) (2.23.0)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.7/dist-packages (from pooch>=1.0->librosa->noisereduce) (20.9)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa->noisereduce) (57.0.0)\n",
            "Requirement already satisfied: llvmlite<0.35,>=0.34.0.dev0 in /usr/local/lib/python3.7/dist-packages (from numba>=0.43.0->librosa->noisereduce) (0.34.0)\n",
            "Requirement already satisfied: cffi>=1.0 in /usr/local/lib/python3.7/dist-packages (from soundfile>=0.9.0->librosa->noisereduce) (1.14.5)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->noisereduce) (2.10)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->noisereduce) (3.0.4)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->noisereduce) (1.24.3)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->pooch>=1.0->librosa->noisereduce) (2020.12.5)\n",
            "Requirement already satisfied: pycparser in /usr/local/lib/python3.7/dist-packages (from cffi>=1.0->soundfile>=0.9.0->librosa->noisereduce) (2.20)\n",
            "Building wheels for collected packages: noisereduce\n",
            "  Building wheel for noisereduce (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for noisereduce: filename=noisereduce-1.1.0-cp37-none-any.whl size=7610 sha256=d9f8c6966c0a58b459a1a06d3acd8c7249bb4e6dc662a7eb3a3da3a9fc30793c\n",
            "  Stored in directory: /root/.cache/pip/wheels/a6/2c/70/f9ccb41280dcfbe3eaeb7930f913dd85394617f3d3136f29cc\n",
            "Successfully built noisereduce\n",
            "Installing collected packages: noisereduce\n",
            "Successfully installed noisereduce-1.1.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "T3KOtxybHYl2",
        "outputId": "6cf750d3-41a0-4fee-c999-8d5893a36e3d"
      },
      "source": [
        "import os,sys \n",
        "import re\n",
        "sys.path.append('/content/gdrive/My Drive/birdclef-2021')\n",
        "path = ('/content/gdrive/My Drive/birdclef-2021/')\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings(action='ignore')\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "import seaborn as sns\n",
        "from tqdm import tqdm\n",
        "import pickle\n",
        "\n",
        "import librosa\n",
        "import noisereduce\n",
        "import librosa.display\n",
        "import IPython.display as ipd\n",
        "\n",
        "from sklearn.utils import shuffle\n",
        "from PIL import Image\n",
        "\n",
        "import torch\n",
        "from torchvision import transforms\n",
        "\n",
        "import tensorflow as tf\n",
        "from tensorflow.keras.applications import ResNet50\n",
        "from tensorflow.keras.applications import ResNet101V2\n",
        "from tensorflow.keras.applications import ResNet152V2\n",
        "\n",
        "import keras\n",
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras.models import Model, load_model\n",
        "#from keras.initializers import glorot_uniform\n",
        "from tensorflow.keras.layers import Input, Dropout, Add, Dense\n",
        "from tensorflow.keras.layers import AveragePooling2D, Reshape, Activation\n",
        "from tensorflow.keras.layers import BatchNormalization, Flatten, Conv1D\n",
        "\n",
        "import utils\n",
        "\n",
        "# Global vars\n",
        "RANDOM_SEED = 1337\n",
        "SAMPLE_RATE = 32000\n",
        "SIGNAL_LENGTH = 5 # seconds\n",
        "SPEC_SHAPE = (48, 128) # height x width\n",
        "INPUT_SHAPE = SPEC_SHAPE + (3,)\n",
        "FMIN = 500\n",
        "FMAX = 12500\n",
        "MAX_AUDIO_FILES = 1500\n",
        "\n",
        "input_dir = path + 'basic_test/birds/'\n",
        "output_dir = path + 'basic_test/melspectrogram_dataset/'\n",
        "model_dir = path + 'basic_test/'\n",
        "soundscape_dir = path + 'basic_test/'"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Hello Kaare\n",
            "utils\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "PXOkch3tHYmD"
      },
      "source": [
        "## Get data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-BLP84TYHYmH"
      },
      "source": [
        "### Select data\n",
        "Find all bird species with more than 200 entries in the training data and save that data as TRAIN and LABELS"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "137SrGM6HYmN",
        "outputId": "f7d8cf8c-1959-4601-ce80-e9acb8bcfa89"
      },
      "source": [
        "# Load metadata file\n",
        "train = pd.read_csv(path + 'train_metadata.csv',)\n",
        "\n",
        "# Limit the number of training samples and classes\n",
        "# First, only use high quality samples\n",
        "train = train.query('rating>=4')\n",
        "\n",
        "# Second, assume that birds with the most training samples are also the most common\n",
        "# A species needs at least 200 recordings with a rating above 4 to be considered common\n",
        "birds_count = {}\n",
        "for bird_species, count in zip(train.primary_label.unique(), \n",
        "                               train.groupby('primary_label')['primary_label'].count().values):\n",
        "    birds_count[bird_species] = count\n",
        "most_represented_birds = [key for key,value in birds_count.items() if value >= 200] \n",
        "\n",
        "TRAIN = train.query('primary_label in @most_represented_birds')\n",
        "LABELS = sorted(TRAIN.primary_label.unique())\n",
        "\n",
        "# Let's see how many species and samples we have left\n",
        "print('NUMBER OF SPECIES IN TRAIN DATA:', len(LABELS))\n",
        "print('NUMBER OF SAMPLES IN TRAIN DATA:', len(TRAIN))\n",
        "print('LABELS:', most_represented_birds)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "NUMBER OF SPECIES IN TRAIN DATA: 27\n",
            "NUMBER OF SAMPLES IN TRAIN DATA: 8548\n",
            "LABELS: ['amerob', 'barswa', 'bewwre', 'blujay', 'bncfly', 'carwre', 'compau', 'comrav', 'comyel', 'eursta', 'gbwwre1', 'grekis', 'houspa', 'houwre', 'mallar3', 'norcar', 'normoc', 'redcro', 'rewbla', 'roahaw', 'rubpep1', 'rucspa1', 'sonspa', 'spotow', 'wbwwre1', 'wesmea', 'yeofly1']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rMax8TeNHYmU",
        "outputId": "777c138e-747e-417e-f8d7-060401168d45"
      },
      "source": [
        "# Shuffle the training data and limit the number of audio files to MAX_AUDIO_FILES\n",
        "TRAIN = shuffle(TRAIN, random_state=RANDOM_SEED)[:MAX_AUDIO_FILES]\n",
        "\n",
        "print('FINAL NUMBER OF AUDIO FILES IN TRAINING DATA:', len(TRAIN))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "FINAL NUMBER OF AUDIO FILES IN TRAINING DATA: 1500\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pP1PC--FHYmY"
      },
      "source": [
        "### Load spectograms\n",
        "Run the load function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "MI0RXMt0HYmc",
        "outputId": "ab2afb3d-d69c-417e-b1ec-9d6a8c597e03"
      },
      "source": [
        "# Parse audio files and extract training samples\n",
        "samples = []\n",
        "with tqdm(total=len(TRAIN)) as pbar:\n",
        "    for idx, row in TRAIN.iterrows():\n",
        "        pbar.update(1)\n",
        "        \n",
        "        if row.primary_label in most_represented_birds:\n",
        "            audio_file_path = os.path.join(input_dir, row.primary_label, row.filename)\n",
        "            samples += utils.get_spectrograms(audio_file_path, row.primary_label, output_dir, SAMPLE_RATE, SIGNAL_LENGTH, SPEC_SHAPE, FMIN, FMAX)\n",
        "            \n",
        "TRAIN_SPECS = shuffle(samples, random_state=RANDOM_SEED)\n",
        "print('SUCCESSFULLY EXTRACTED {} SPECTROGRAMS'.format(len(TRAIN_SPECS)))\n",
        "\n",
        "with open(path + \"samples_paths.txt\", \"wb\") as fp:\n",
        "  pickle.dump(TRAIN_SPECS, fp)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "100%|██████████| 1500/1500 [38:44<00:00,  1.55s/it]\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "SUCCESSFULLY EXTRACTED 4157 SPECTROGRAMS\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yQ5xAxqxHYmw"
      },
      "source": [
        "### Load data\n",
        "Load spectograms and normalize, and add to a stack of 4154 sound bites, of 48 (MEL resolution) x 128 (time resolution) pixels. Labels are 4154 one-hot vectors of length 27 (number of bird types)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4K4fi1wE7m_P"
      },
      "source": [
        "with open(path + \"samples_paths.txt\", \"rb\") as fp:\n",
        "  TRAIN_SPECS = pickle.load(fp)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "AQVjC5syHYm9",
        "outputId": "d6ecd9dc-d1e9-4b60-d000-320bc81e6bce"
      },
      "source": [
        "# Parse all samples and add spectrograms into train data, primary_labels into label data\n",
        "train_specs, train_labels = [], []\n",
        "with tqdm(total=len(TRAIN_SPECS)) as pbar:\n",
        "    for path in TRAIN_SPECS:\n",
        "        pbar.update(1)\n",
        "\n",
        "        # Open image\n",
        "        spec = Image.open(path)\n",
        "\n",
        "        # Convert to numpy array\n",
        "        spec = np.array(spec, dtype='float32')\n",
        "        \n",
        "        # Normalize between 0.0 and 1.0\n",
        "        # and exclude samples with nan \n",
        "        spec -= spec.min()\n",
        "        spec /= spec.max()\n",
        "        if not spec.max() == 1.0 or not spec.min() == 0.0:\n",
        "            continue\n",
        "\n",
        "        # Add channel axis to 2D array\n",
        "        spec = np.expand_dims(spec, -1)\n",
        "\n",
        "        # Add new dimension for batch size\n",
        "        spec = np.expand_dims(spec, 0)\n",
        "\n",
        "        # Add to train data\n",
        "        if len(train_specs) == 0:\n",
        "            train_specs = spec\n",
        "        else:\n",
        "            train_specs = np.vstack((train_specs, spec))\n",
        "\n",
        "        # Add to label data\n",
        "        target = np.zeros((len(LABELS)), dtype='float32')\n",
        "        bird = path.replace(os.sep, '/').split('/')[-2]\n",
        "        target[LABELS.index(bird)] = 1.0\n",
        "        if len(train_labels) == 0:\n",
        "            train_labels = target\n",
        "        else:\n",
        "            train_labels = np.vstack((train_labels, target))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            " 36%|███▌      | 1485/4157 [10:02<17:08,  2.60it/s]"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Eowg3DBTHYnL"
      },
      "source": [
        "### Repeat in channel dimension and preprocess"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ydjmPkXJHYnR"
      },
      "source": [
        "train_input = tf.keras.applications.resnet.preprocess_input(np.repeat(train_specs, 3, axis=3))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "vIa_ijCLHYnV"
      },
      "source": [
        "## Create model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nz77qTvIbBqI",
        "outputId": "fa235cb6-a4c8-49ca-a2dd-3b6abff8260f"
      },
      "source": [
        "!nvcc --version"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "nvcc: NVIDIA (R) Cuda compiler driver\n",
            "Copyright (c) 2005-2020 NVIDIA Corporation\n",
            "Built on Wed_Jul_22_19:09:09_PDT_2020\n",
            "Cuda compilation tools, release 11.0, V11.0.221\n",
            "Build cuda_11.0_bu.TC445_37.28845127_0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NCVvryIuPZX4"
      },
      "source": [
        "tf.device('/device:GPU:0')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "tags": [],
        "id": "iOa7L5zrHYnb"
      },
      "source": [
        "### Load base model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RYaydQVaHYnf"
      },
      "source": [
        "baseModel = ResNet50(weights='imagenet', include_top=False, input_tensor=Input(shape=(INPUT_SHAPE)))\n",
        "#baseModel = ResNet101V2(weights='imagenet', include_top=False, input_tensor=Input(shape=(INPUT_SHAPE)))\n",
        "#baseModel = ResNet152V2(weights='imagenet', include_top=False, input_tensor=Input(shape=(INPUT_SHAPE)))\n",
        "\n",
        "train_input = tf.keras.applications.resnet.preprocess_input(np.repeat(train_specs, 3, axis=3))\n",
        "#train_input = tf.keras.applications.resnet_v2.preprocess_input(np.repeat(train_specs, 3, axis=3))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TWo1qCcRHYnv",
        "outputId": "c5fd21d6-23a5-4aa4-9095-ec5a99ef0121"
      },
      "source": [
        "baseModel.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"resnet50\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 48, 128, 3)] 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv1_pad (ZeroPadding2D)       (None, 54, 134, 3)   0           input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "conv1_conv (Conv2D)             (None, 24, 64, 64)   9472        conv1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv1_bn (BatchNormalization)   (None, 24, 64, 64)   256         conv1_conv[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv1_relu (Activation)         (None, 24, 64, 64)   0           conv1_bn[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "pool1_pad (ZeroPadding2D)       (None, 26, 66, 64)   0           conv1_relu[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "pool1_pool (MaxPooling2D)       (None, 12, 32, 64)   0           pool1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_conv (Conv2D)    (None, 12, 32, 64)   4160        pool1_pool[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_relu (Activation (None, 12, 32, 64)   0           conv2_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_relu (Activation (None, 12, 32, 64)   0           conv2_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_0_conv (Conv2D)    (None, 12, 32, 256)  16640       pool1_pool[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_0_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_add (Add)          (None, 12, 32, 256)  0           conv2_block1_0_bn[0][0]          \n",
            "                                                                 conv2_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_out (Activation)   (None, 12, 32, 256)  0           conv2_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_conv (Conv2D)    (None, 12, 32, 64)   16448       conv2_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_relu (Activation (None, 12, 32, 64)   0           conv2_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_relu (Activation (None, 12, 32, 64)   0           conv2_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_add (Add)          (None, 12, 32, 256)  0           conv2_block1_out[0][0]           \n",
            "                                                                 conv2_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_out (Activation)   (None, 12, 32, 256)  0           conv2_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_conv (Conv2D)    (None, 12, 32, 64)   16448       conv2_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_relu (Activation (None, 12, 32, 64)   0           conv2_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_relu (Activation (None, 12, 32, 64)   0           conv2_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_add (Add)          (None, 12, 32, 256)  0           conv2_block2_out[0][0]           \n",
            "                                                                 conv2_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_out (Activation)   (None, 12, 32, 256)  0           conv2_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_conv (Conv2D)    (None, 6, 16, 128)   32896       conv2_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_relu (Activation (None, 6, 16, 128)   0           conv3_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_relu (Activation (None, 6, 16, 128)   0           conv3_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_0_conv (Conv2D)    (None, 6, 16, 512)   131584      conv2_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_0_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_add (Add)          (None, 6, 16, 512)   0           conv3_block1_0_bn[0][0]          \n",
            "                                                                 conv3_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_out (Activation)   (None, 6, 16, 512)   0           conv3_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_relu (Activation (None, 6, 16, 128)   0           conv3_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_relu (Activation (None, 6, 16, 128)   0           conv3_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_add (Add)          (None, 6, 16, 512)   0           conv3_block1_out[0][0]           \n",
            "                                                                 conv3_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_out (Activation)   (None, 6, 16, 512)   0           conv3_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_relu (Activation (None, 6, 16, 128)   0           conv3_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_relu (Activation (None, 6, 16, 128)   0           conv3_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_add (Add)          (None, 6, 16, 512)   0           conv3_block2_out[0][0]           \n",
            "                                                                 conv3_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_out (Activation)   (None, 6, 16, 512)   0           conv3_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block4_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_relu (Activation (None, 6, 16, 128)   0           conv3_block4_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block4_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block4_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_relu (Activation (None, 6, 16, 128)   0           conv3_block4_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block4_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block4_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_add (Add)          (None, 6, 16, 512)   0           conv3_block3_out[0][0]           \n",
            "                                                                 conv3_block4_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_out (Activation)   (None, 6, 16, 512)   0           conv3_block4_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_conv (Conv2D)    (None, 3, 8, 256)    131328      conv3_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_relu (Activation (None, 3, 8, 256)    0           conv4_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_relu (Activation (None, 3, 8, 256)    0           conv4_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_0_conv (Conv2D)    (None, 3, 8, 1024)   525312      conv3_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_0_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_add (Add)          (None, 3, 8, 1024)   0           conv4_block1_0_bn[0][0]          \n",
            "                                                                 conv4_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_out (Activation)   (None, 3, 8, 1024)   0           conv4_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_relu (Activation (None, 3, 8, 256)    0           conv4_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_relu (Activation (None, 3, 8, 256)    0           conv4_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_add (Add)          (None, 3, 8, 1024)   0           conv4_block1_out[0][0]           \n",
            "                                                                 conv4_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_out (Activation)   (None, 3, 8, 1024)   0           conv4_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_relu (Activation (None, 3, 8, 256)    0           conv4_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_relu (Activation (None, 3, 8, 256)    0           conv4_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_add (Add)          (None, 3, 8, 1024)   0           conv4_block2_out[0][0]           \n",
            "                                                                 conv4_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_out (Activation)   (None, 3, 8, 1024)   0           conv4_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block4_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_relu (Activation (None, 3, 8, 256)    0           conv4_block4_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block4_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block4_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_relu (Activation (None, 3, 8, 256)    0           conv4_block4_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block4_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block4_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_add (Add)          (None, 3, 8, 1024)   0           conv4_block3_out[0][0]           \n",
            "                                                                 conv4_block4_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_out (Activation)   (None, 3, 8, 1024)   0           conv4_block4_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block5_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_relu (Activation (None, 3, 8, 256)    0           conv4_block5_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block5_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block5_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_relu (Activation (None, 3, 8, 256)    0           conv4_block5_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block5_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block5_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_add (Add)          (None, 3, 8, 1024)   0           conv4_block4_out[0][0]           \n",
            "                                                                 conv4_block5_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_out (Activation)   (None, 3, 8, 1024)   0           conv4_block5_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block5_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block6_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_relu (Activation (None, 3, 8, 256)    0           conv4_block6_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block6_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block6_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_relu (Activation (None, 3, 8, 256)    0           conv4_block6_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block6_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block6_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_add (Add)          (None, 3, 8, 1024)   0           conv4_block5_out[0][0]           \n",
            "                                                                 conv4_block6_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_out (Activation)   (None, 3, 8, 1024)   0           conv4_block6_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_conv (Conv2D)    (None, 2, 4, 512)    524800      conv4_block6_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_relu (Activation (None, 2, 4, 512)    0           conv5_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_relu (Activation (None, 2, 4, 512)    0           conv5_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_0_conv (Conv2D)    (None, 2, 4, 2048)   2099200     conv4_block6_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_0_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_add (Add)          (None, 2, 4, 2048)   0           conv5_block1_0_bn[0][0]          \n",
            "                                                                 conv5_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_out (Activation)   (None, 2, 4, 2048)   0           conv5_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_conv (Conv2D)    (None, 2, 4, 512)    1049088     conv5_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_relu (Activation (None, 2, 4, 512)    0           conv5_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_relu (Activation (None, 2, 4, 512)    0           conv5_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_add (Add)          (None, 2, 4, 2048)   0           conv5_block1_out[0][0]           \n",
            "                                                                 conv5_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_out (Activation)   (None, 2, 4, 2048)   0           conv5_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_conv (Conv2D)    (None, 2, 4, 512)    1049088     conv5_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_relu (Activation (None, 2, 4, 512)    0           conv5_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_relu (Activation (None, 2, 4, 512)    0           conv5_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_add (Add)          (None, 2, 4, 2048)   0           conv5_block2_out[0][0]           \n",
            "                                                                 conv5_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_out (Activation)   (None, 2, 4, 2048)   0           conv5_block3_add[0][0]           \n",
            "==================================================================================================\n",
            "Total params: 23,587,712\n",
            "Trainable params: 23,534,592\n",
            "Non-trainable params: 53,120\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sy5pyiLoHYn2"
      },
      "source": [
        "### Create a new head"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L9fpK4JTHYn6"
      },
      "source": [
        "headModel = baseModel.output\n",
        "headModel = AveragePooling2D(pool_size=(2, 4))(headModel)\n",
        "#headModel = AveragePooling2D(pool_size=(7, 7))(headModel)\n",
        "headModel = Flatten(name=\"flatten\")(headModel)\n",
        "\n",
        "headModel = Dense(256, activation=\"relu\")(headModel)\n",
        "headModel = Dropout(0.5)(headModel)\n",
        "\n",
        "headModel = Dense(128, activation=\"relu\")(headModel)\n",
        "headModel = Dropout(0.5)(headModel)\n",
        "\n",
        "headModel = Dense(64, activation=\"relu\")(headModel)\n",
        "headModel = Dropout(0.5)(headModel)\n",
        "\n",
        "headModel = Dense(len(LABELS), activation=\"softmax\")(headModel)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lbw2a28uHYn7"
      },
      "source": [
        "### Add to model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "tags": [],
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9ibbN9z4HYn-",
        "outputId": "f1a9d32c-a94e-481c-ad32-b5844b7153b5"
      },
      "source": [
        "model = Model(inputs=baseModel.input, outputs=headModel)\n",
        "\n",
        "# Reset trainability\n",
        "for layer in baseModel.layers:\n",
        "\tlayer.trainable = True\n",
        "\n",
        "model.summary()\n",
        "\n",
        "# Select untrainable layers\n",
        "#for layer in baseModel.layers[12:-24]:\n",
        "\t#layer.trainable = False\n",
        "\n",
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 48, 128, 3)] 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv1_pad (ZeroPadding2D)       (None, 54, 134, 3)   0           input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "conv1_conv (Conv2D)             (None, 24, 64, 64)   9472        conv1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv1_bn (BatchNormalization)   (None, 24, 64, 64)   256         conv1_conv[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv1_relu (Activation)         (None, 24, 64, 64)   0           conv1_bn[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "pool1_pad (ZeroPadding2D)       (None, 26, 66, 64)   0           conv1_relu[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "pool1_pool (MaxPooling2D)       (None, 12, 32, 64)   0           pool1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_conv (Conv2D)    (None, 12, 32, 64)   4160        pool1_pool[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_relu (Activation (None, 12, 32, 64)   0           conv2_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_relu (Activation (None, 12, 32, 64)   0           conv2_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_0_conv (Conv2D)    (None, 12, 32, 256)  16640       pool1_pool[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_0_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_add (Add)          (None, 12, 32, 256)  0           conv2_block1_0_bn[0][0]          \n",
            "                                                                 conv2_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_out (Activation)   (None, 12, 32, 256)  0           conv2_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_conv (Conv2D)    (None, 12, 32, 64)   16448       conv2_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_relu (Activation (None, 12, 32, 64)   0           conv2_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_relu (Activation (None, 12, 32, 64)   0           conv2_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_add (Add)          (None, 12, 32, 256)  0           conv2_block1_out[0][0]           \n",
            "                                                                 conv2_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_out (Activation)   (None, 12, 32, 256)  0           conv2_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_conv (Conv2D)    (None, 12, 32, 64)   16448       conv2_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_relu (Activation (None, 12, 32, 64)   0           conv2_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_relu (Activation (None, 12, 32, 64)   0           conv2_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_add (Add)          (None, 12, 32, 256)  0           conv2_block2_out[0][0]           \n",
            "                                                                 conv2_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_out (Activation)   (None, 12, 32, 256)  0           conv2_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_conv (Conv2D)    (None, 6, 16, 128)   32896       conv2_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_relu (Activation (None, 6, 16, 128)   0           conv3_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_relu (Activation (None, 6, 16, 128)   0           conv3_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_0_conv (Conv2D)    (None, 6, 16, 512)   131584      conv2_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_0_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_add (Add)          (None, 6, 16, 512)   0           conv3_block1_0_bn[0][0]          \n",
            "                                                                 conv3_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_out (Activation)   (None, 6, 16, 512)   0           conv3_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_relu (Activation (None, 6, 16, 128)   0           conv3_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_relu (Activation (None, 6, 16, 128)   0           conv3_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_add (Add)          (None, 6, 16, 512)   0           conv3_block1_out[0][0]           \n",
            "                                                                 conv3_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_out (Activation)   (None, 6, 16, 512)   0           conv3_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_relu (Activation (None, 6, 16, 128)   0           conv3_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_relu (Activation (None, 6, 16, 128)   0           conv3_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_add (Add)          (None, 6, 16, 512)   0           conv3_block2_out[0][0]           \n",
            "                                                                 conv3_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_out (Activation)   (None, 6, 16, 512)   0           conv3_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block4_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_relu (Activation (None, 6, 16, 128)   0           conv3_block4_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block4_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block4_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_relu (Activation (None, 6, 16, 128)   0           conv3_block4_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block4_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block4_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_add (Add)          (None, 6, 16, 512)   0           conv3_block3_out[0][0]           \n",
            "                                                                 conv3_block4_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_out (Activation)   (None, 6, 16, 512)   0           conv3_block4_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_conv (Conv2D)    (None, 3, 8, 256)    131328      conv3_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_relu (Activation (None, 3, 8, 256)    0           conv4_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_relu (Activation (None, 3, 8, 256)    0           conv4_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_0_conv (Conv2D)    (None, 3, 8, 1024)   525312      conv3_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_0_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_add (Add)          (None, 3, 8, 1024)   0           conv4_block1_0_bn[0][0]          \n",
            "                                                                 conv4_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_out (Activation)   (None, 3, 8, 1024)   0           conv4_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_relu (Activation (None, 3, 8, 256)    0           conv4_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_relu (Activation (None, 3, 8, 256)    0           conv4_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_add (Add)          (None, 3, 8, 1024)   0           conv4_block1_out[0][0]           \n",
            "                                                                 conv4_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_out (Activation)   (None, 3, 8, 1024)   0           conv4_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_relu (Activation (None, 3, 8, 256)    0           conv4_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_relu (Activation (None, 3, 8, 256)    0           conv4_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_add (Add)          (None, 3, 8, 1024)   0           conv4_block2_out[0][0]           \n",
            "                                                                 conv4_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_out (Activation)   (None, 3, 8, 1024)   0           conv4_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block4_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_relu (Activation (None, 3, 8, 256)    0           conv4_block4_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block4_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block4_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_relu (Activation (None, 3, 8, 256)    0           conv4_block4_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block4_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block4_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_add (Add)          (None, 3, 8, 1024)   0           conv4_block3_out[0][0]           \n",
            "                                                                 conv4_block4_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_out (Activation)   (None, 3, 8, 1024)   0           conv4_block4_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block5_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_relu (Activation (None, 3, 8, 256)    0           conv4_block5_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block5_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block5_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_relu (Activation (None, 3, 8, 256)    0           conv4_block5_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block5_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block5_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_add (Add)          (None, 3, 8, 1024)   0           conv4_block4_out[0][0]           \n",
            "                                                                 conv4_block5_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_out (Activation)   (None, 3, 8, 1024)   0           conv4_block5_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block5_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block6_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_relu (Activation (None, 3, 8, 256)    0           conv4_block6_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block6_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block6_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_relu (Activation (None, 3, 8, 256)    0           conv4_block6_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block6_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block6_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_add (Add)          (None, 3, 8, 1024)   0           conv4_block5_out[0][0]           \n",
            "                                                                 conv4_block6_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_out (Activation)   (None, 3, 8, 1024)   0           conv4_block6_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_conv (Conv2D)    (None, 2, 4, 512)    524800      conv4_block6_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_relu (Activation (None, 2, 4, 512)    0           conv5_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_relu (Activation (None, 2, 4, 512)    0           conv5_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_0_conv (Conv2D)    (None, 2, 4, 2048)   2099200     conv4_block6_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_0_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_add (Add)          (None, 2, 4, 2048)   0           conv5_block1_0_bn[0][0]          \n",
            "                                                                 conv5_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_out (Activation)   (None, 2, 4, 2048)   0           conv5_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_conv (Conv2D)    (None, 2, 4, 512)    1049088     conv5_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_relu (Activation (None, 2, 4, 512)    0           conv5_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_relu (Activation (None, 2, 4, 512)    0           conv5_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_add (Add)          (None, 2, 4, 2048)   0           conv5_block1_out[0][0]           \n",
            "                                                                 conv5_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_out (Activation)   (None, 2, 4, 2048)   0           conv5_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_conv (Conv2D)    (None, 2, 4, 512)    1049088     conv5_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_relu (Activation (None, 2, 4, 512)    0           conv5_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_relu (Activation (None, 2, 4, 512)    0           conv5_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_add (Add)          (None, 2, 4, 2048)   0           conv5_block2_out[0][0]           \n",
            "                                                                 conv5_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_out (Activation)   (None, 2, 4, 2048)   0           conv5_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d (AveragePooli (None, 1, 1, 2048)   0           conv5_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "flatten (Flatten)               (None, 2048)         0           average_pooling2d[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "dense (Dense)                   (None, 256)          524544      flatten[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout (Dropout)               (None, 256)          0           dense[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 128)          32896       dropout[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout_1 (Dropout)             (None, 128)          0           dense_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dense_2 (Dense)                 (None, 64)           8256        dropout_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dropout_2 (Dropout)             (None, 64)           0           dense_2[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dense_3 (Dense)                 (None, 27)           1755        dropout_2[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 24,155,163\n",
            "Trainable params: 24,102,043\n",
            "Non-trainable params: 53,120\n",
            "__________________________________________________________________________________________________\n",
            "Model: \"model\"\n",
            "__________________________________________________________________________________________________\n",
            "Layer (type)                    Output Shape         Param #     Connected to                     \n",
            "==================================================================================================\n",
            "input_1 (InputLayer)            [(None, 48, 128, 3)] 0                                            \n",
            "__________________________________________________________________________________________________\n",
            "conv1_pad (ZeroPadding2D)       (None, 54, 134, 3)   0           input_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "conv1_conv (Conv2D)             (None, 24, 64, 64)   9472        conv1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv1_bn (BatchNormalization)   (None, 24, 64, 64)   256         conv1_conv[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv1_relu (Activation)         (None, 24, 64, 64)   0           conv1_bn[0][0]                   \n",
            "__________________________________________________________________________________________________\n",
            "pool1_pad (ZeroPadding2D)       (None, 26, 66, 64)   0           conv1_relu[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "pool1_pool (MaxPooling2D)       (None, 12, 32, 64)   0           pool1_pad[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_conv (Conv2D)    (None, 12, 32, 64)   4160        pool1_pool[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_1_relu (Activation (None, 12, 32, 64)   0           conv2_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_2_relu (Activation (None, 12, 32, 64)   0           conv2_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_0_conv (Conv2D)    (None, 12, 32, 256)  16640       pool1_pool[0][0]                 \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_0_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_add (Add)          (None, 12, 32, 256)  0           conv2_block1_0_bn[0][0]          \n",
            "                                                                 conv2_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block1_out (Activation)   (None, 12, 32, 256)  0           conv2_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_conv (Conv2D)    (None, 12, 32, 64)   16448       conv2_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_1_relu (Activation (None, 12, 32, 64)   0           conv2_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_2_relu (Activation (None, 12, 32, 64)   0           conv2_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_add (Add)          (None, 12, 32, 256)  0           conv2_block1_out[0][0]           \n",
            "                                                                 conv2_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block2_out (Activation)   (None, 12, 32, 256)  0           conv2_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_conv (Conv2D)    (None, 12, 32, 64)   16448       conv2_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_1_relu (Activation (None, 12, 32, 64)   0           conv2_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_conv (Conv2D)    (None, 12, 32, 64)   36928       conv2_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_bn (BatchNormali (None, 12, 32, 64)   256         conv2_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_2_relu (Activation (None, 12, 32, 64)   0           conv2_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_3_conv (Conv2D)    (None, 12, 32, 256)  16640       conv2_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_3_bn (BatchNormali (None, 12, 32, 256)  1024        conv2_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_add (Add)          (None, 12, 32, 256)  0           conv2_block2_out[0][0]           \n",
            "                                                                 conv2_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv2_block3_out (Activation)   (None, 12, 32, 256)  0           conv2_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_conv (Conv2D)    (None, 6, 16, 128)   32896       conv2_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_1_relu (Activation (None, 6, 16, 128)   0           conv3_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_2_relu (Activation (None, 6, 16, 128)   0           conv3_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_0_conv (Conv2D)    (None, 6, 16, 512)   131584      conv2_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_0_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_add (Add)          (None, 6, 16, 512)   0           conv3_block1_0_bn[0][0]          \n",
            "                                                                 conv3_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block1_out (Activation)   (None, 6, 16, 512)   0           conv3_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_1_relu (Activation (None, 6, 16, 128)   0           conv3_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_2_relu (Activation (None, 6, 16, 128)   0           conv3_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_add (Add)          (None, 6, 16, 512)   0           conv3_block1_out[0][0]           \n",
            "                                                                 conv3_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block2_out (Activation)   (None, 6, 16, 512)   0           conv3_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_1_relu (Activation (None, 6, 16, 128)   0           conv3_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_2_relu (Activation (None, 6, 16, 128)   0           conv3_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_add (Add)          (None, 6, 16, 512)   0           conv3_block2_out[0][0]           \n",
            "                                                                 conv3_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block3_out (Activation)   (None, 6, 16, 512)   0           conv3_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_conv (Conv2D)    (None, 6, 16, 128)   65664       conv3_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block4_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_1_relu (Activation (None, 6, 16, 128)   0           conv3_block4_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_conv (Conv2D)    (None, 6, 16, 128)   147584      conv3_block4_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_bn (BatchNormali (None, 6, 16, 128)   512         conv3_block4_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_2_relu (Activation (None, 6, 16, 128)   0           conv3_block4_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_3_conv (Conv2D)    (None, 6, 16, 512)   66048       conv3_block4_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_3_bn (BatchNormali (None, 6, 16, 512)   2048        conv3_block4_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_add (Add)          (None, 6, 16, 512)   0           conv3_block3_out[0][0]           \n",
            "                                                                 conv3_block4_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv3_block4_out (Activation)   (None, 6, 16, 512)   0           conv3_block4_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_conv (Conv2D)    (None, 3, 8, 256)    131328      conv3_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_1_relu (Activation (None, 3, 8, 256)    0           conv4_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_2_relu (Activation (None, 3, 8, 256)    0           conv4_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_0_conv (Conv2D)    (None, 3, 8, 1024)   525312      conv3_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_0_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_add (Add)          (None, 3, 8, 1024)   0           conv4_block1_0_bn[0][0]          \n",
            "                                                                 conv4_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block1_out (Activation)   (None, 3, 8, 1024)   0           conv4_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_1_relu (Activation (None, 3, 8, 256)    0           conv4_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_2_relu (Activation (None, 3, 8, 256)    0           conv4_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_add (Add)          (None, 3, 8, 1024)   0           conv4_block1_out[0][0]           \n",
            "                                                                 conv4_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block2_out (Activation)   (None, 3, 8, 1024)   0           conv4_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_1_relu (Activation (None, 3, 8, 256)    0           conv4_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_2_relu (Activation (None, 3, 8, 256)    0           conv4_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_add (Add)          (None, 3, 8, 1024)   0           conv4_block2_out[0][0]           \n",
            "                                                                 conv4_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block3_out (Activation)   (None, 3, 8, 1024)   0           conv4_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block4_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_1_relu (Activation (None, 3, 8, 256)    0           conv4_block4_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block4_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block4_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_2_relu (Activation (None, 3, 8, 256)    0           conv4_block4_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block4_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block4_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_add (Add)          (None, 3, 8, 1024)   0           conv4_block3_out[0][0]           \n",
            "                                                                 conv4_block4_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block4_out (Activation)   (None, 3, 8, 1024)   0           conv4_block4_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block4_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block5_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_1_relu (Activation (None, 3, 8, 256)    0           conv4_block5_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block5_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block5_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_2_relu (Activation (None, 3, 8, 256)    0           conv4_block5_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block5_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block5_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_add (Add)          (None, 3, 8, 1024)   0           conv4_block4_out[0][0]           \n",
            "                                                                 conv4_block5_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block5_out (Activation)   (None, 3, 8, 1024)   0           conv4_block5_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_conv (Conv2D)    (None, 3, 8, 256)    262400      conv4_block5_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block6_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_1_relu (Activation (None, 3, 8, 256)    0           conv4_block6_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_conv (Conv2D)    (None, 3, 8, 256)    590080      conv4_block6_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_bn (BatchNormali (None, 3, 8, 256)    1024        conv4_block6_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_2_relu (Activation (None, 3, 8, 256)    0           conv4_block6_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_3_conv (Conv2D)    (None, 3, 8, 1024)   263168      conv4_block6_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_3_bn (BatchNormali (None, 3, 8, 1024)   4096        conv4_block6_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_add (Add)          (None, 3, 8, 1024)   0           conv4_block5_out[0][0]           \n",
            "                                                                 conv4_block6_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv4_block6_out (Activation)   (None, 3, 8, 1024)   0           conv4_block6_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_conv (Conv2D)    (None, 2, 4, 512)    524800      conv4_block6_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block1_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_1_relu (Activation (None, 2, 4, 512)    0           conv5_block1_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block1_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block1_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_2_relu (Activation (None, 2, 4, 512)    0           conv5_block1_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_0_conv (Conv2D)    (None, 2, 4, 2048)   2099200     conv4_block6_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block1_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_0_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block1_0_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block1_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_add (Add)          (None, 2, 4, 2048)   0           conv5_block1_0_bn[0][0]          \n",
            "                                                                 conv5_block1_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block1_out (Activation)   (None, 2, 4, 2048)   0           conv5_block1_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_conv (Conv2D)    (None, 2, 4, 512)    1049088     conv5_block1_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block2_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_1_relu (Activation (None, 2, 4, 512)    0           conv5_block2_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block2_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block2_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_2_relu (Activation (None, 2, 4, 512)    0           conv5_block2_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block2_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block2_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_add (Add)          (None, 2, 4, 2048)   0           conv5_block1_out[0][0]           \n",
            "                                                                 conv5_block2_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block2_out (Activation)   (None, 2, 4, 2048)   0           conv5_block2_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_conv (Conv2D)    (None, 2, 4, 512)    1049088     conv5_block2_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block3_1_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_1_relu (Activation (None, 2, 4, 512)    0           conv5_block3_1_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_conv (Conv2D)    (None, 2, 4, 512)    2359808     conv5_block3_1_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_bn (BatchNormali (None, 2, 4, 512)    2048        conv5_block3_2_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_2_relu (Activation (None, 2, 4, 512)    0           conv5_block3_2_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_3_conv (Conv2D)    (None, 2, 4, 2048)   1050624     conv5_block3_2_relu[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_3_bn (BatchNormali (None, 2, 4, 2048)   8192        conv5_block3_3_conv[0][0]        \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_add (Add)          (None, 2, 4, 2048)   0           conv5_block2_out[0][0]           \n",
            "                                                                 conv5_block3_3_bn[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "conv5_block3_out (Activation)   (None, 2, 4, 2048)   0           conv5_block3_add[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "average_pooling2d (AveragePooli (None, 1, 1, 2048)   0           conv5_block3_out[0][0]           \n",
            "__________________________________________________________________________________________________\n",
            "flatten (Flatten)               (None, 2048)         0           average_pooling2d[0][0]          \n",
            "__________________________________________________________________________________________________\n",
            "dense (Dense)                   (None, 256)          524544      flatten[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout (Dropout)               (None, 256)          0           dense[0][0]                      \n",
            "__________________________________________________________________________________________________\n",
            "dense_1 (Dense)                 (None, 128)          32896       dropout[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dropout_1 (Dropout)             (None, 128)          0           dense_1[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dense_2 (Dense)                 (None, 64)           8256        dropout_1[0][0]                  \n",
            "__________________________________________________________________________________________________\n",
            "dropout_2 (Dropout)             (None, 64)           0           dense_2[0][0]                    \n",
            "__________________________________________________________________________________________________\n",
            "dense_3 (Dense)                 (None, 27)           1755        dropout_2[0][0]                  \n",
            "==================================================================================================\n",
            "Total params: 24,155,163\n",
            "Trainable params: 24,102,043\n",
            "Non-trainable params: 53,120\n",
            "__________________________________________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "o0zjBYzrbJmC",
        "outputId": "04f3e15c-3c9e-45d7-9fb2-7d4ee53c5311"
      },
      "source": [
        "n_layers = 0\n",
        "for layer in baseModel.layers:\n",
        "  n_layers +=1\n",
        "  print(n_layers, layer.name)\n",
        "  print(layer.trainable)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1 input_1\n",
            "True\n",
            "2 conv1_pad\n",
            "True\n",
            "3 conv1_conv\n",
            "True\n",
            "4 conv1_bn\n",
            "True\n",
            "5 conv1_relu\n",
            "True\n",
            "6 pool1_pad\n",
            "True\n",
            "7 pool1_pool\n",
            "True\n",
            "8 conv2_block1_1_conv\n",
            "True\n",
            "9 conv2_block1_1_bn\n",
            "True\n",
            "10 conv2_block1_1_relu\n",
            "True\n",
            "11 conv2_block1_2_conv\n",
            "True\n",
            "12 conv2_block1_2_bn\n",
            "True\n",
            "13 conv2_block1_2_relu\n",
            "True\n",
            "14 conv2_block1_0_conv\n",
            "True\n",
            "15 conv2_block1_3_conv\n",
            "True\n",
            "16 conv2_block1_0_bn\n",
            "True\n",
            "17 conv2_block1_3_bn\n",
            "True\n",
            "18 conv2_block1_add\n",
            "True\n",
            "19 conv2_block1_out\n",
            "True\n",
            "20 conv2_block2_1_conv\n",
            "True\n",
            "21 conv2_block2_1_bn\n",
            "True\n",
            "22 conv2_block2_1_relu\n",
            "True\n",
            "23 conv2_block2_2_conv\n",
            "True\n",
            "24 conv2_block2_2_bn\n",
            "True\n",
            "25 conv2_block2_2_relu\n",
            "True\n",
            "26 conv2_block2_3_conv\n",
            "True\n",
            "27 conv2_block2_3_bn\n",
            "True\n",
            "28 conv2_block2_add\n",
            "True\n",
            "29 conv2_block2_out\n",
            "True\n",
            "30 conv2_block3_1_conv\n",
            "True\n",
            "31 conv2_block3_1_bn\n",
            "True\n",
            "32 conv2_block3_1_relu\n",
            "True\n",
            "33 conv2_block3_2_conv\n",
            "True\n",
            "34 conv2_block3_2_bn\n",
            "True\n",
            "35 conv2_block3_2_relu\n",
            "True\n",
            "36 conv2_block3_3_conv\n",
            "True\n",
            "37 conv2_block3_3_bn\n",
            "True\n",
            "38 conv2_block3_add\n",
            "True\n",
            "39 conv2_block3_out\n",
            "True\n",
            "40 conv3_block1_1_conv\n",
            "True\n",
            "41 conv3_block1_1_bn\n",
            "True\n",
            "42 conv3_block1_1_relu\n",
            "True\n",
            "43 conv3_block1_2_conv\n",
            "True\n",
            "44 conv3_block1_2_bn\n",
            "True\n",
            "45 conv3_block1_2_relu\n",
            "True\n",
            "46 conv3_block1_0_conv\n",
            "True\n",
            "47 conv3_block1_3_conv\n",
            "True\n",
            "48 conv3_block1_0_bn\n",
            "True\n",
            "49 conv3_block1_3_bn\n",
            "True\n",
            "50 conv3_block1_add\n",
            "True\n",
            "51 conv3_block1_out\n",
            "True\n",
            "52 conv3_block2_1_conv\n",
            "True\n",
            "53 conv3_block2_1_bn\n",
            "True\n",
            "54 conv3_block2_1_relu\n",
            "True\n",
            "55 conv3_block2_2_conv\n",
            "True\n",
            "56 conv3_block2_2_bn\n",
            "True\n",
            "57 conv3_block2_2_relu\n",
            "True\n",
            "58 conv3_block2_3_conv\n",
            "True\n",
            "59 conv3_block2_3_bn\n",
            "True\n",
            "60 conv3_block2_add\n",
            "True\n",
            "61 conv3_block2_out\n",
            "True\n",
            "62 conv3_block3_1_conv\n",
            "True\n",
            "63 conv3_block3_1_bn\n",
            "True\n",
            "64 conv3_block3_1_relu\n",
            "True\n",
            "65 conv3_block3_2_conv\n",
            "True\n",
            "66 conv3_block3_2_bn\n",
            "True\n",
            "67 conv3_block3_2_relu\n",
            "True\n",
            "68 conv3_block3_3_conv\n",
            "True\n",
            "69 conv3_block3_3_bn\n",
            "True\n",
            "70 conv3_block3_add\n",
            "True\n",
            "71 conv3_block3_out\n",
            "True\n",
            "72 conv3_block4_1_conv\n",
            "True\n",
            "73 conv3_block4_1_bn\n",
            "True\n",
            "74 conv3_block4_1_relu\n",
            "True\n",
            "75 conv3_block4_2_conv\n",
            "True\n",
            "76 conv3_block4_2_bn\n",
            "True\n",
            "77 conv3_block4_2_relu\n",
            "True\n",
            "78 conv3_block4_3_conv\n",
            "True\n",
            "79 conv3_block4_3_bn\n",
            "True\n",
            "80 conv3_block4_add\n",
            "True\n",
            "81 conv3_block4_out\n",
            "True\n",
            "82 conv4_block1_1_conv\n",
            "True\n",
            "83 conv4_block1_1_bn\n",
            "True\n",
            "84 conv4_block1_1_relu\n",
            "True\n",
            "85 conv4_block1_2_conv\n",
            "True\n",
            "86 conv4_block1_2_bn\n",
            "True\n",
            "87 conv4_block1_2_relu\n",
            "True\n",
            "88 conv4_block1_0_conv\n",
            "True\n",
            "89 conv4_block1_3_conv\n",
            "True\n",
            "90 conv4_block1_0_bn\n",
            "True\n",
            "91 conv4_block1_3_bn\n",
            "True\n",
            "92 conv4_block1_add\n",
            "True\n",
            "93 conv4_block1_out\n",
            "True\n",
            "94 conv4_block2_1_conv\n",
            "True\n",
            "95 conv4_block2_1_bn\n",
            "True\n",
            "96 conv4_block2_1_relu\n",
            "True\n",
            "97 conv4_block2_2_conv\n",
            "True\n",
            "98 conv4_block2_2_bn\n",
            "True\n",
            "99 conv4_block2_2_relu\n",
            "True\n",
            "100 conv4_block2_3_conv\n",
            "True\n",
            "101 conv4_block2_3_bn\n",
            "True\n",
            "102 conv4_block2_add\n",
            "True\n",
            "103 conv4_block2_out\n",
            "True\n",
            "104 conv4_block3_1_conv\n",
            "True\n",
            "105 conv4_block3_1_bn\n",
            "True\n",
            "106 conv4_block3_1_relu\n",
            "True\n",
            "107 conv4_block3_2_conv\n",
            "True\n",
            "108 conv4_block3_2_bn\n",
            "True\n",
            "109 conv4_block3_2_relu\n",
            "True\n",
            "110 conv4_block3_3_conv\n",
            "True\n",
            "111 conv4_block3_3_bn\n",
            "True\n",
            "112 conv4_block3_add\n",
            "True\n",
            "113 conv4_block3_out\n",
            "True\n",
            "114 conv4_block4_1_conv\n",
            "True\n",
            "115 conv4_block4_1_bn\n",
            "True\n",
            "116 conv4_block4_1_relu\n",
            "True\n",
            "117 conv4_block4_2_conv\n",
            "True\n",
            "118 conv4_block4_2_bn\n",
            "True\n",
            "119 conv4_block4_2_relu\n",
            "True\n",
            "120 conv4_block4_3_conv\n",
            "True\n",
            "121 conv4_block4_3_bn\n",
            "True\n",
            "122 conv4_block4_add\n",
            "True\n",
            "123 conv4_block4_out\n",
            "True\n",
            "124 conv4_block5_1_conv\n",
            "True\n",
            "125 conv4_block5_1_bn\n",
            "True\n",
            "126 conv4_block5_1_relu\n",
            "True\n",
            "127 conv4_block5_2_conv\n",
            "True\n",
            "128 conv4_block5_2_bn\n",
            "True\n",
            "129 conv4_block5_2_relu\n",
            "True\n",
            "130 conv4_block5_3_conv\n",
            "True\n",
            "131 conv4_block5_3_bn\n",
            "True\n",
            "132 conv4_block5_add\n",
            "True\n",
            "133 conv4_block5_out\n",
            "True\n",
            "134 conv4_block6_1_conv\n",
            "True\n",
            "135 conv4_block6_1_bn\n",
            "True\n",
            "136 conv4_block6_1_relu\n",
            "True\n",
            "137 conv4_block6_2_conv\n",
            "True\n",
            "138 conv4_block6_2_bn\n",
            "True\n",
            "139 conv4_block6_2_relu\n",
            "True\n",
            "140 conv4_block6_3_conv\n",
            "True\n",
            "141 conv4_block6_3_bn\n",
            "True\n",
            "142 conv4_block6_add\n",
            "True\n",
            "143 conv4_block6_out\n",
            "True\n",
            "144 conv5_block1_1_conv\n",
            "True\n",
            "145 conv5_block1_1_bn\n",
            "True\n",
            "146 conv5_block1_1_relu\n",
            "True\n",
            "147 conv5_block1_2_conv\n",
            "True\n",
            "148 conv5_block1_2_bn\n",
            "True\n",
            "149 conv5_block1_2_relu\n",
            "True\n",
            "150 conv5_block1_0_conv\n",
            "True\n",
            "151 conv5_block1_3_conv\n",
            "True\n",
            "152 conv5_block1_0_bn\n",
            "True\n",
            "153 conv5_block1_3_bn\n",
            "True\n",
            "154 conv5_block1_add\n",
            "True\n",
            "155 conv5_block1_out\n",
            "True\n",
            "156 conv5_block2_1_conv\n",
            "True\n",
            "157 conv5_block2_1_bn\n",
            "True\n",
            "158 conv5_block2_1_relu\n",
            "True\n",
            "159 conv5_block2_2_conv\n",
            "True\n",
            "160 conv5_block2_2_bn\n",
            "True\n",
            "161 conv5_block2_2_relu\n",
            "True\n",
            "162 conv5_block2_3_conv\n",
            "True\n",
            "163 conv5_block2_3_bn\n",
            "True\n",
            "164 conv5_block2_add\n",
            "True\n",
            "165 conv5_block2_out\n",
            "True\n",
            "166 conv5_block3_1_conv\n",
            "True\n",
            "167 conv5_block3_1_bn\n",
            "True\n",
            "168 conv5_block3_1_relu\n",
            "True\n",
            "169 conv5_block3_2_conv\n",
            "True\n",
            "170 conv5_block3_2_bn\n",
            "True\n",
            "171 conv5_block3_2_relu\n",
            "True\n",
            "172 conv5_block3_3_conv\n",
            "True\n",
            "173 conv5_block3_3_bn\n",
            "True\n",
            "174 conv5_block3_add\n",
            "True\n",
            "175 conv5_block3_out\n",
            "True\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "frdDACC1HYoH"
      },
      "source": [
        "model.compile(optimizer=tf.keras.optimizers.Adam(lr=0.001),\n",
        "              loss=tf.keras.losses.CategoricalCrossentropy(label_smoothing=0.01),\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QlcyELq-HYoS"
      },
      "source": [
        "# Add callbacks to reduce the learning rate if needed, early stopping, and checkpoint saving\n",
        "callbacks = [tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', \n",
        "                                                  patience=2, \n",
        "                                                  verbose=1, \n",
        "                                                  factor=0.5),\n",
        "             tf.keras.callbacks.EarlyStopping(monitor='val_loss', \n",
        "                                              verbose=1,\n",
        "                                              patience=5),\n",
        "             tf.keras.callbacks.ModelCheckpoint(filepath=model_dir + 'best_model.h5', \n",
        "                                                monitor='val_loss',\n",
        "                                                verbose=0,\n",
        "                                                save_best_only=True)]\n",
        "\n",
        "callbacks = [tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', \n",
        "                                                  patience=2, \n",
        "                                                  verbose=1, \n",
        "                                                  factor=0.5),\n",
        "             tf.keras.callbacks.ModelCheckpoint(filepath=model_dir + 'best_model.h5', \n",
        "                                                monitor='val_loss',\n",
        "                                                verbose=0,\n",
        "                                                save_best_only=True)]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yQ2TDT9OHYoV"
      },
      "source": [
        "## Train model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8xfY1Y2UHYoa",
        "outputId": "1da439db-3b75-4bf8-da4a-8a00397d7214"
      },
      "source": [
        "model.fit(train_input, \n",
        "          train_labels,\n",
        "          batch_size=32,\n",
        "          validation_split=0.2,\n",
        "          callbacks=callbacks,\n",
        "          epochs=20)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/20\n",
            "104/104 [==============================] - 56s 195ms/step - loss: 3.3478 - accuracy: 0.0536 - val_loss: 3.2746 - val_accuracy: 0.0710\n",
            "Epoch 2/20\n",
            "104/104 [==============================] - 18s 171ms/step - loss: 3.2756 - accuracy: 0.0686 - val_loss: 3.2580 - val_accuracy: 0.0710\n",
            "Epoch 3/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2616 - accuracy: 0.0701 - val_loss: 3.2383 - val_accuracy: 0.0710\n",
            "Epoch 4/20\n",
            "104/104 [==============================] - 18s 170ms/step - loss: 3.2531 - accuracy: 0.0668 - val_loss: 3.2254 - val_accuracy: 0.0710\n",
            "Epoch 5/20\n",
            "104/104 [==============================] - 18s 170ms/step - loss: 3.2453 - accuracy: 0.0674 - val_loss: 3.2157 - val_accuracy: 0.0710\n",
            "Epoch 6/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2420 - accuracy: 0.0680 - val_loss: 3.2115 - val_accuracy: 0.0710\n",
            "Epoch 7/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2399 - accuracy: 0.0641 - val_loss: 3.2103 - val_accuracy: 0.0710\n",
            "Epoch 8/20\n",
            "104/104 [==============================] - 18s 170ms/step - loss: 3.2445 - accuracy: 0.0620 - val_loss: 3.2106 - val_accuracy: 0.0710\n",
            "Epoch 9/20\n",
            "104/104 [==============================] - 17s 167ms/step - loss: 3.2381 - accuracy: 0.0647 - val_loss: 3.2102 - val_accuracy: 0.0710\n",
            "Epoch 10/20\n",
            "104/104 [==============================] - 17s 168ms/step - loss: 3.2399 - accuracy: 0.0662 - val_loss: 3.2096 - val_accuracy: 0.0710\n",
            "Epoch 11/20\n",
            "104/104 [==============================] - 18s 170ms/step - loss: 3.2400 - accuracy: 0.0686 - val_loss: 3.2089 - val_accuracy: 0.0710\n",
            "Epoch 12/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2394 - accuracy: 0.0662 - val_loss: 3.2088 - val_accuracy: 0.0710\n",
            "Epoch 13/20\n",
            "104/104 [==============================] - 18s 171ms/step - loss: 3.2392 - accuracy: 0.0674 - val_loss: 3.2086 - val_accuracy: 0.0710\n",
            "Epoch 14/20\n",
            "104/104 [==============================] - 18s 170ms/step - loss: 3.2379 - accuracy: 0.0692 - val_loss: 3.2076 - val_accuracy: 0.0710\n",
            "Epoch 15/20\n",
            "104/104 [==============================] - 18s 171ms/step - loss: 3.2388 - accuracy: 0.0641 - val_loss: 3.2085 - val_accuracy: 0.0710\n",
            "Epoch 16/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2382 - accuracy: 0.0707 - val_loss: 3.2082 - val_accuracy: 0.0710\n",
            "\n",
            "Epoch 00016: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\n",
            "Epoch 17/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2377 - accuracy: 0.0683 - val_loss: 3.2082 - val_accuracy: 0.0710\n",
            "Epoch 18/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2387 - accuracy: 0.0689 - val_loss: 3.2078 - val_accuracy: 0.0710\n",
            "\n",
            "Epoch 00018: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\n",
            "Epoch 19/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2370 - accuracy: 0.0659 - val_loss: 3.2077 - val_accuracy: 0.0710\n",
            "Epoch 20/20\n",
            "104/104 [==============================] - 18s 169ms/step - loss: 3.2367 - accuracy: 0.0650 - val_loss: 3.2077 - val_accuracy: 0.0710\n",
            "\n",
            "Epoch 00020: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7fb65bac4d10>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MKKnseLOWVgz"
      },
      "source": [
        "## Homemade"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j4yX3MtfHYoe"
      },
      "source": [
        "def identity_block(X, f, filters, stage, block):\n",
        "    # defining name basis\n",
        "    conv_name_base = 'res' + str(stage) + block + '_branch'\n",
        "    bn_name_base = 'bn' + str(stage) + block + '_branch'\n",
        "\n",
        "    # Retrieve Filters\n",
        "    F1, F2, F3 = filters\n",
        "\n",
        "    # Save the input value. We'll need this later to add back to the main path. \n",
        "    X_shortcut = X\n",
        "\n",
        "    # First component of main path\n",
        "    X = Conv2D(filters = F1, kernel_size = (1, 1), strides = (1,1), padding = 'valid', name = conv_name_base + '2a', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis = 3, name = bn_name_base + '2a')(X)\n",
        "    X = Activation('relu')(X)\n",
        "\n",
        "    # Second component of main path\n",
        "    X = Conv2D(filters = F2, kernel_size = (f, f), strides = (1,1), padding = 'same', name = conv_name_base + '2b', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis = 3, name = bn_name_base + '2b')(X)\n",
        "    X = Activation('relu')(X)\n",
        "\n",
        "    # Third component of main path\n",
        "    X = Conv2D(filters = F3, kernel_size = (1, 1), strides = (1,1), padding = 'valid', name = conv_name_base + '2c', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis = 3, name = bn_name_base + '2c')(X)\n",
        "\n",
        "    # Final step: Add shortcut value to main path, and pass it through a RELU activation\n",
        "    X = Add()([X, X_shortcut])\n",
        "    X = Activation('relu')(X)\n",
        "\n",
        "    return X"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Xxgz6FSCWZSE"
      },
      "source": [
        "def convolutional_block(X, f, filters, stage, block, s = 2):\n",
        "    # defining name basis\n",
        "    conv_name_base = 'res' + str(stage) + block + '_branch'\n",
        "    bn_name_base = 'bn' + str(stage) + block + '_branch'\n",
        "    \n",
        "    # Retrieve Filters\n",
        "    F1, F2, F3 = filters\n",
        "    \n",
        "    # Save the input value\n",
        "    X_shortcut = X\n",
        "\n",
        "\n",
        "    ##### MAIN PATH #####\n",
        "    # First component of main path \n",
        "    X = Conv2D(F1, (1, 1), strides = (s,s), name = conv_name_base + '2a', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis = 3, name = bn_name_base + '2a')(X)\n",
        "    X = Activation('relu')(X)\n",
        "\n",
        "    # Second component of main path\n",
        "    X = Conv2D(filters=F2, kernel_size=(f, f), strides=(1, 1), padding='same', name=conv_name_base + '2b', kernel_initializer=glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis=3, name=bn_name_base + '2b')(X)\n",
        "    X = Activation('relu')(X)\n",
        "\n",
        "    # Third component of main path\n",
        "    X = Conv2D(filters=F3, kernel_size=(1, 1), strides=(1, 1), padding='valid', name=conv_name_base + '2c', kernel_initializer=glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis=3, name=bn_name_base + '2c')(X)\n",
        "\n",
        "    \n",
        "    ##### SHORTCUT PATH ####\n",
        "    X_shortcut = Conv2D(F3, (1, 1), strides = (s,s), name = conv_name_base + '1', kernel_initializer = glorot_uniform(seed=0))(X_shortcut)\n",
        "    X_shortcut = BatchNormalization(axis = 3, name = bn_name_base + '1')(X_shortcut)\n",
        "\n",
        "    # Final step: Add shortcut value to main path, and pass it through a RELU activation\n",
        "    X = Add()([X, X_shortcut])\n",
        "    X = Activation('relu')(X)\n",
        "    \n",
        "    return X"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "h1ap2PdzWdft"
      },
      "source": [
        "def ResNet18(input_shape = (INPUT_SHAPE), classes = len(LABELS)):   \n",
        "    # Define the input as a tensor with shape input_shape\n",
        "    X_input = Input(input_shape)\n",
        "\n",
        "    # Zero-Padding\n",
        "    X = ZeroPadding2D((3, 3))(X_input)\n",
        "    \n",
        "    # Stage 1\n",
        "    X = Conv2D(64, (7, 7), strides = (2, 2), name = 'conv1', kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "    X = BatchNormalization(axis = 3, name = 'bn_conv1')(X)\n",
        "    X = Activation('relu')(X)\n",
        "    X = MaxPooling2D((3, 3), strides=(2, 2))(X)\n",
        "\n",
        "    # Stage 2\n",
        "    X = convolutional_block(X, f = 3, filters = [64, 64, 256], stage = 2, block='a', s = 1)\n",
        "    X = identity_block(X, 3, [64, 64, 256], stage=2, block='b')\n",
        "\n",
        "    # Stage 3\n",
        "    X = convolutional_block(X, f = 3, filters = [128, 128, 512], stage = 3, block='a', s = 2)\n",
        "    X = identity_block(X, 3, [128, 128, 512], stage=3, block='b')\n",
        "\n",
        "\n",
        "    # Stage 4\n",
        "    X = convolutional_block(X, f = 3, filters = [256, 256, 1024], stage = 4, block='a', s = 2)\n",
        "    X = identity_block(X, 3, [256, 256, 1024], stage=4, block='b')\n",
        "\n",
        "\n",
        "    # Stage 5\n",
        "    X = convolutional_block(X, f = 3, filters = [512, 512, 2048], stage = 5, block='a', s = 2)\n",
        "    X = identity_block(X, 3, [512, 512, 2048], stage=5, block='b')\n",
        "\n",
        "    # AVGPOOL.\n",
        "    X = AveragePooling2D((2, 2), name='avg_pool')(X)\n",
        "\n",
        "    # output layer\n",
        "    X = Flatten()(X)\n",
        "\n",
        "    X = Dense(256, activation=\"relu\")(X)\n",
        "    X = Dropout(0.5)(X)\n",
        "\n",
        "    X = Dense(256, activation=\"relu\")(X)\n",
        "    X = Dropout(0.5)(X)\n",
        "\n",
        "    X = Dense(classes, activation='softmax', name='fc' + str(classes), kernel_initializer = glorot_uniform(seed=0))(X)\n",
        "    \n",
        "    # Create model\n",
        "    model = Model(inputs = X_input, outputs = X, name='ResNet50')\n",
        "\n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gMGb4NYpYHbo"
      },
      "source": [
        "os.environ['CUDA_VISIBLE_DEVICES'] = '0'\n",
        "import cv2\n",
        "import numpy as np\n",
        "from keras import layers\n",
        "from keras.layers import Input, Add, Dense, Activation, ZeroPadding2D, BatchNormalization, Flatten, Conv2D, AveragePooling2D, MaxPooling2D\n",
        "from keras.models import Model, load_model\n",
        "from keras.initializers import glorot_uniform\n",
        "from tensorflow.keras.utils import plot_model\n",
        "from IPython.display import SVG\n",
        "from keras.utils.vis_utils import model_to_dot\n",
        "import keras.backend as K\n",
        "import tensorflow as tf\n",
        "\n",
        "\n",
        "\n",
        "model = ResNet18(input_shape = INPUT_SHAPE, classes = len(LABELS))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n-jVAiKsYRRn"
      },
      "source": [
        "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eKNxJs_JYYEP"
      },
      "source": [
        "model.fit(train_input, \n",
        "          train_labels,\n",
        "          batch_size=32,\n",
        "          validation_split=0.2,\n",
        "          callbacks=callbacks,\n",
        "          epochs=50)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Pk0eACY7YrFo"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}