FaVOS / README.md
FaVOSsubmission's picture
Upload README.md with huggingface_hub
04b47c7 verified
---
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.