Single reference
Bistnet
Paper: https://github.com/yyang181/NTIRE23-VIDEO-COLORIZATION
This method is originally designed for video colorization and is limited to 2 reference images. The model is changed so that the input is 2 identical frames. The flow map is set to zeros and the similarity fusion is done using argmax for an arbitrary number of reference images.
| Task | Image #1 | Image #2 | Image #3 | Image #4 |
|---|---|---|---|---|
| Recolor source |   |
  |
  |
  |
| Inverted reference |   |
  |
  |
  |
| Task | Image #1 | Image #2 | Image #3 | Reference |
| Full correspondence |   |
  |
  |
  |
| Full correspondence |   |
  |
  |
  |
| Partial reference |   |
  |
  |
  |
| Partial reference |   |
  |
  |
  |
| Partial source |   |
  |
  |
|
| Partial source |   |
  |
  |
|
| Semantic correspondence strong |   |
  |
  |
|
| Semantic correspondence strong |   |
  |
  |
|
| Semantic correspondence weak |   |
  |
  |
  |
| Semantic correspondence weak |   |
  |
  |
  |
| Distractors |   |
  |
  |
|
| Random noise |   |
  |
  |
  |
| Random noise |   |
  |
  |
|
| Gray |   |
  |
  |
  |
| Gray |   |
  |
  |
  |
Additional Information
- Last updated: 10 April 2024 10:47
- GPU info: NVIDIA GeForce GTX 1080 Ti 11 GB, Compute Capability 6.1
- CUDA version: 11.7
- PyTorch version: 1.13.1