PyTorch implementation of ESPCN [1]/VESPCN [2].
- Add your own template in template.py, indicating parameters related to running the code (especially, specify the task (Image/MC/Video) and set training/test dataset directories specific to your filesystem)
- Add your model in ./model/ directory (filename should be in lower cases)
- Type "python3 main.py --template
$(your template) --model $ (model you want to train)" for training - If you want to add additional options for test benchmark datasets, modify ./data/init.py.
- For additional details, refer to [3] (We have borrowed most of the implementation details from there).
- Implement the SISR ESPCN network
- Making dataloader for video SR
- Complete the motion compensation network
- Joining the ESPCN to motion compensation network
[1] W. Shi et al, “Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network,” IEEE CVPR 2016.
[2] J. Caballero et al, “Real-Time Video Super-Resolution with Spatio-Temporal Networks and Motion Compensation,” IEEE CVPR 2017.
[3] https://github.com/thstkdgus35/EDSR-PyTorch (borrowed the overall code structure)