peterzhousz / starnet

Official Implementation of Space-Time-Aware Multi-Resolution Video Enhancement (CVPR2020) using PyTorch

Space-Time-Aware Multi-Resolution Video Enhancement (CVPR2020)

Project page: https://alterzero.github.io/projects/STAR.html

Dependencies

• Python 3.5
• PyTorch >= 1.0.0
• Pyflow -> https://github.com/pathak22/pyflow

  cd [STARnet]
  git clone https://github.com/pathak22/pyflow
  cd pyflow/
  python setup.py build_ext -i
  cp pyflow*.so ..

Dataset

• Vimeo-90k triplets Dataset
• UCF101
• Middlebury dataset

Pretrained Model and Testset

https://drive.google.com/drive/folders/1cLZhlx4PLhF75qIZYxuVKLRKzN5_6i8r?usp=sharing

HOW TO

Train w/o FR

   python    main.py    

Train w/ FR

   python    main_refinement_flow.py    

Train w/ FR for STAR-T-HR

   python   main_refinement_t_sr_hr.py    

Evaluate STAR, STAR-ST, STAR-S, STAR-T-LR

   python    eval.py    

Evaluate STAR-T-HR

   python    eval_star_t_hr.py    

Related work

Deep Back-Projection Networks for Super-Resolution (CVPR2018)

Project page: https://alterzero.github.io/projects/DBPN.html

Winner (1st) of NTIRE2018 Competition (Track: x8 Bicubic Downsampling)

Winner of PIRM2018 (1st on Region 2, 3rd on Region 1, and 5th on Region 3)

Recurrent Back-Projection Network for Video Super-Resolution (CVPR2019)

Project page: https://alterzero.github.io/projects/RBPN.html

Honorable mention of NTIRE2019 (Video Super-resolution and Deblurring)

Citations

If you find this work useful, please consider citing it.

@inproceedings{STAR2020,
  title={Space-Time-Aware Multi-Resolution Video Enhancement},
  author={Haris, Muhammad and Shakhnarovich, Greg and Ukita, Norimichi},
  booktitle={IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2020}
}