- Tensorboard Support
- GIF
SPMC_VideoSR
├── cfgs
│ ├── config_latest.py
│ └── config.py
├── data
│ └── test
│ └── 1.jpg
├── dataset.py
├── losses.py
├── model.py
├── modules
│ ├── detail_fusion_net.py
│ ├── __init__.py
│ ├── me.py
│ └── spmc.py
├── np.py
├── predict.py
├── preprocess.py
├── utils
│ ├── flow.py
│ ├── color.py
│ ├──
└── WORKBENCH.ipynb- Preprocess
- Train
- Predict
- Load model
| Input | Output | Groundtruth |
|---|---|---|
 |
 |
 |
(PSNR/SSIM)
| Method | Ours | Paper |
|---|---|---|
| ×2 | 36.71 / 0.96 | |
| ×3 | 31.92 / 0.90 | |
| ×4 | 29.69 / 0.84 |
- Our flow visualization is different from MPI-Sintel's. for more details about flow estimation.
- Use ffmpeg to make GIFs
- Paper: Detail-revealing Deep Video Super-resolution
- Offical Repo: jiangsutx/SPMC_VideoSR