In this project, we use Autoencoder and GAN to tackle this problem. We train our models on DIV2K dataset and evaluate the performance on JPEG AI public test dataset based on PSNR and SSIM.
python3 train.py --dataDir path_to_your_training_data
python3 test.py --dataDir path_to_your_testing_data --encoderWeight ./checkpoint/encoder.pt --decoderWeight ./checkpoint/decoder.pt