metadata
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:
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:
data/Annotations/<video_id>/<frame>.png
Use favos-20.txt and favos-40.txt as the evaluation splits.