- To install dependences:
Ideally create a new conda enviroment with python==3.6
Then install a GPU-enabled pytorch==1.6.0, torchvision=0.7.0
Finally from the conda env, run the extra dependencies if they are not already satified
pip install -r pip_requirements.txt
- To train and validate the model per epoch run:
python train.py --model_dir=experiments/base_model --data_dir=/path/to/Dataset
Default value for model_dir is `experiments/base_model`
Change the `/path/to/Dataset` to the correct path of the ArtNet
Also, to change the hyper-parameters, refer to the params.json file since everything is controlled from there.
- To run inference on the model:
python inference.py --model_dir=experiments/base_model --data_dir=/path/to/Dataset -- restore_from=/path/to/best_weights/checkpoint.pth.tar
Change the `/path/to/Dataset` to the correct path of the ArtNet
Change the `/path/to/best_weights/checkpoint.pth.tar` to the correct path of the pre-trained checkpoint (the default value is `experiments/best_model/best_weights/best_after_epoch_163.pth.tar`)
Finally, if you used any of the ARTGAN code, please cite the following paper:
"A. Adhikary, N. Bhandari, E. Markou and S. Sachan, "ArtGAN: Artwork Restoration using Generative Adversarial Networks," 2021 13th International Conference on Advanced Computational Intelligence (ICACI), 2021, pp. 199-206, doi: 10.1109/ICACI52617.2021.9435888."