| license: cc-by-nc-4.0 | |
| tags: | |
| - video-object-segmentation | |
| - video-segmentation | |
| - evaluation | |
| # FaVOS | |
| FaVOS evaluation uses the first annotated frame as the prompt frame. Metrics are | |
| computed on the remaining frames only. | |
| Download the dataset: | |
| ```bash | |
| huggingface-cli download FaVOSsubmission/FaVOS --repo-type dataset --local-dir data | |
| unzip data/JPEGImages.zip -d data | |
| unzip data/Annotations.zip -d data | |
| ``` | |
| Dataset files are available at: | |
| https://huggingface.co/datasets/FaVOSsubmission/FaVOS/tree/main | |
| Evaluate predictions against: | |
| ```text | |
| data/Annotations/<video_id>/<frame>.png | |
| ``` | |
| Use `favos-20.txt` and `favos-40.txt` as the evaluation splits. | |