PyTorch implementation of this paper with minor variations for my deep learning course final project.
Add dataset pathes to params.py, you can also change the network and optimization parameters in this file. To start training the model or resume from last saved checkpoint, run
python main.py --train
To evaluate the model on test or validation set run
python main.py --evaluate --evalset [test or validation] --checkpoint path/to/checkpoint
if you don't provide --checkpoint argument, it will use last saved checkpoint.