From 6a9dfadd09b1ab4344ca5c15bea58c7c0b808f5e Mon Sep 17 00:00:00 2001 From: MlgJcqt <41188079+MlgJcqt@users.noreply.github.com> Date: Tue, 10 Nov 2020 12:44:08 +0100 Subject: [PATCH] adapt compare.py to compare images from 2 folders unresolved issue : load&align phase really slow for folders above 5000 images --- contributed/Facenet1vN | 204 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 204 insertions(+) create mode 100644 contributed/Facenet1vN diff --git a/contributed/Facenet1vN b/contributed/Facenet1vN new file mode 100644 index 000000000..037daa734 --- /dev/null +++ b/contributed/Facenet1vN @@ -0,0 +1,204 @@ +# author : Maelig Jacquet +# June 2020 +# adapted from davidsanberg/facenet + +"""Performs face alignment and calculates L2 distance between the embeddings of images from 2 folders.""" + +# MIT License +# +# Copyright (c) 2016 David Sandberg +# +# Permission is hereby granted, free of charge, to any person obtaining a copy +# of this software and associated documentation files (the "Software"), to deal +# in the Software without restriction, including without limitation the rights +# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +# copies of the Software, and to permit persons to whom the Software is +# furnished to do so, subject to the following conditions: +# +# The above copyright notice and this permission notice shall be included in all +# copies or substantial portions of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +# SOFTWARE. + +from __future__ import absolute_import +from __future__ import division +from __future__ import print_function + +from scipy import misc +import tensorflow as tf +import numpy as np +import sys +import os +import copy +import argparse +import facenet +import align.detect_face +import xlsxwriter +import itertools +import cv2 +from tqdm import tqdm +#import imageio + +parser = argparse.ArgumentParser() + + +parser.add_argument('image_files', type=str, nargs='+', help='Images to compare') +parser.add_argument('--image_size', type=int, + help='Image size (height, width) in pixels.', default=160) +parser.add_argument('--margin', type=int, + help='Margin for the crop around the bounding box (height, width) in pixels.', default=44) +parser.add_argument('--gpu_memory_fraction', type=float, + help='Upper bound on the amount of GPU memory that will be used by the process.', default=0.01) +parser.add_argument('--model', type=str, + help='Could be either a directory containing the meta_file and ckpt_file or a model protobuf (.pb) file', default='../data/model/20180402-114759.pb') +parser.add_argument('out_file', type=str, + help='Output file.', default='../output/results.csv') + +args = parser.parse_args() + +a_same = args.a_same +a_different = args.a_different +b_same = args.b_same +b_different = args.b_different +outfile = args.out_file + +dirimg1 = args.image_files[0] +dirimg2 = args.image_files[1] + +if os.path.isdir(dirimg1): + listimg1 = [] + for f in os.listdir(dirimg1): + listimg1.append(dirimg1 + "/" + f ) +else : + listimg1=[dirimg1] + + +if os.path.isdir(dirimg2): + listimg2 = [] + for f in os.listdir(dirimg2): + listimg2.append(dirimg2 + "/" + f ) +else : + listimg2=[dirimg2] + + +def main(listimg1, listimg2): + with tf.Graph().as_default(): + + with tf.Session() as sess: + + # Load the model + facenet.load_model(args.model) + + # Get input and output tensors + images_placeholder = tf.get_default_graph().get_tensor_by_name("input:0") + embeddings = tf.get_default_graph().get_tensor_by_name("embeddings:0") + phase_train_placeholder = tf.get_default_graph().get_tensor_by_name("phase_train:0") + + # Run forward pass to calculate embeddings + ###### original compare.py modif + nrof_traces = len(listimg1) + nrof_bdd = len(listimg2) + + emb1 = np.array([]) + emb2 = np.array([]) + + feed_dict1 = { images_placeholder: images1, phase_train_placeholder:False } + emb1 = sess.run(embeddings, feed_dict=feed_dict1) + + feed_dict2 = { images_placeholder: images2, phase_train_placeholder:False } + emb2 = sess.run(embeddings, feed_dict=feed_dict2) + + print (nrof_traces, "images in folder 1") + print (nrof_bdd, "images in folder 2") + print (nrof_traces * nrof_bdd, "comparisons to run") + + # Create output folder if doesnt exist + out_path = os.path.dirname(outfile) + if not os.path.exists(out_path): + os.makedirs(out_path) + + outf = open(outfile, "w+") + outf.write("Image 1;Image 2;Score\n") + + print ('MATCHING...') + + for (i, a) in zip(range(len(images1)), listimg1): + nom_img1 = os.path.basename(a) + for (j, b) in zip(range(len(images2)), listimg2): + nom_img2 = os.path.basename(b) + dist = np.sqrt(np.sum(np.square(np.subtract(emb1[i,:], emb2[j,:])))) + + outf.write("%s;%s;%.3f\n" % (nom_img1, nom_img2, dist)) + + outf.close() + + + print ('END') + + +def load_and_align_data(image_paths, image_size, margin, gpu_memory_fraction): + + minsize = 20 # minimum size of face + threshold = [ 0.6, 0.7, 0.7 ] # three steps's threshold + factor = 0.709 # scale factor + + + images1 = [] + images2 = [] + + print('Creating networks and loading parameters for :', (os.path.basename(os.path.dirname(image_paths[0])))) + with tf.Graph().as_default(): + gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=gpu_memory_fraction) + sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options, log_device_placement=False)) + with sess.as_default(): + pnet, rnet, onet = align.detect_face.create_mtcnn(sess, None) + + with tqdm(total=len(image_paths), desc = 'Loading and aligning images from %s' % (os.path.basename(os.path.dirname(image_paths[0])))) as pbar2: + img_list = [] + for image in image_paths: + img = misc.imread(os.path.expanduser(image), mode='RGB') + + img_size = np.asarray(img.shape)[0:2] + bounding_boxes, _ = align.detect_face.detect_face(img, minsize, pnet, rnet, onet, threshold, factor) + if len(bounding_boxes) < 1: + # image_paths.remove(image) + print("can't detect face, remove ", image) + if image in image_paths: + image_paths.remove(image) + continue + det = np.squeeze(bounding_boxes[0,0:4]) + bb = np.zeros(4, dtype=np.int32) + bb[0] = np.maximum(det[0]-margin/2, 0) + bb[1] = np.maximum(det[1]-margin/2, 0) + bb[2] = np.minimum(det[2]+margin/2, img_size[1]) + bb[3] = np.minimum(det[3]+margin/2, img_size[0]) + cropped = img[bb[1]:bb[3],bb[0]:bb[2],:] + aligned = misc.imresize(cropped, (image_size, image_size), interp='bilinear') + prewhitened = facenet.prewhiten(aligned) + + img_list.append(prewhitened) + pbar2.update(1) + + if image_paths == listimg1: + images1 = np.stack(img_list) + + elif image_paths == listimg2: + images2 = np.stack(img_list) + + return images1 if image_paths == listimg1 else images2 + + +images1= load_and_align_data(listimg1, args.image_size, args.margin, args.gpu_memory_fraction) +images2= load_and_align_data(listimg2, args.image_size, args.margin, args.gpu_memory_fraction) + +main(listimg1, listimg2) + +# if __name__ == '__main__': +# main(parse_arguments(sys.argv[1:])) +# print (sys.argv)