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