Pytorch implementation for image compression and reconstruction via autoencoder
This is an autoencoder with cylic loss and coding parsing loss for image compression and reconstruction. Network backbone is simple 3-layer fully conv (encoder) and symmetrical for decoder. Finally it can achieve 21 mean PSNR on CLIC dataset (CVPR 2019 workshop).
You can download training data from this url: https://drive.google.com/drive/folders/1wU1CO6WcQOraIaY2KSk7cRVaAXcm_A2R?usp=sharing
validation data: https://drive.google.com/drive/folders/113EcrAdcxfVqs8BVt4PZjwUEyVz7VVa-?usp=sharing
Organize your data with this structure:
Data | |---train | |---image1.xxx |---image2.xxx . . .
Data_valid | |---train | |---image1.xxx |---image2.xxx . . .
You can train your own model via run_train.sh and modify config as your needs. Prediction for the valid data via run_test.sh