This project aims to detect and crop the relatively sharp patches from an input video. With this module, we can extract the top r% of the sharpest patches from the video. For example, if a video contains 1000 frames and we want to obtain the top 20% of the sharpest patches, we will get 200 patches.
The details will be in our ECCV 2024 Paper. The downstream task is our DADeblur for video deblurring.
dataset/
├── video1/
│ ├── Blur/RGB/
├── video2/
│ ├── Blur/RGB/
├── video3/
│ ├── Blur/RGB/The upstream task is our Blur Magnitude Estimator, you can get the details and download model weight from this repo BME
you need to replace the path in main.py line 114 - 119
python main.py