A user-friendly software for the automatic detection of archaeological features from ALS data using convolutional neural networks. The underlying ML models were trained on an extensive archive of ALS datasets in Ireland, labelled by experts with three types of archaeological features (barrows, ringforts, enclosures). The core components of the tool are the Relief Visualisation Toolbox (RVT) for processing the input data and the Artificial Intelligence Toolbox for Earth Observation (AiTLAS), which provides access to the ML models.
The installation is currently only supported on Windows 64-bit machines. The application is compatible with machines
equipped with CUDA-enabled graphics cards, but will also work on a standard CPU where GPU processing is not possible.
We recommend creating a virtual environment with Anaconda
and installing the requirements with pip
.
See the Step by step instructions below.
-
Python 3.8 (recommended to use conda virtual environment: download and install Miniconda or Anaconda)
-
Files with trained Machine learning models. Download from Dropbox.
Warning: Download contains data with large file size ~5GB in total. This includes 8 pretrained ML models saved as TAR files - 4 for semantic segmentation and 4 for object detection.
-
Clone the repository to your local drive
-
Move the TAR files to
<path-to-repository>\adaf\ml_models
Do not change the filenames and make sure that files are copied to the exact location!
-
Run Anaconda Prompt (press
Windows
key and type “anaconda prompt”). -
In the Anaconda Prompt, navigate to the installation folder by running commands:
cd <path-to-repository> cd installation
<path-to-repository>
is the location where you have downloaded and unzipped the installation files, for exampleC:\temp\adaf\
-
Create and activate a conda environment called
adaf
. Run commands:conda create -n adaf python=3.8 conda activate adaf
-
Install PyTorch for CUDA
Skip this step if you don’t have a CUDA enabled device!
ONLY FOR CUDA COMPLIANT GPUs. When installing on a PC which has a CUDA enabled graphics card (check here for NVIDIA compliant cards) the GPU can be used to reduce processing times. If your card is compliant (also requires installation of CUDA software that is not covered in this manual) install the compatible PyTorch version.
-
Install the packages using pip:
pip install GDAL-3.4.3-cp38-cp38-win_amd64.whl pip install aitlas-0.0.1-py3-none-any.whl
-
Enable the use of the AiTLAS virtual environment in Jupyter notebooks by running:
python -m ipykernel install --name adaf
-
Navigate back to main adaf folder and run Jupyter Notebook with the following command:
cd .. jupyter notebook ADAF_main.ipynb