|
Download README.md from taryya/RunningBench-Annotation-Packs: direct link, hf CLI and curl.
- Browser
- Download file 1.61 kB
-
https://huggingface.co/datasets/taryya/RunningBench-Annotation-Packs/resolve/main/README.md
- Command line
-
hf download hf://datasets/taryya/RunningBench-Annotation-Packs/README.md
-
curl -L -H "Authorization: Bearer $HF_TOKEN" -o README.md https://huggingface.co/datasets/taryya/RunningBench-Annotation-Packs/resolve/main/README.md
1.61 kB
metadata
pretty_name: RunningBench Annotation Packs
language:
- en
- zh
tags:
- video
- annotation
- benchmark
- egocentric-video
RunningBench Annotation Packs
This repository contains ten self-contained offline packages for human verification of 698 RunningBench multiple-choice video questions.
What is included
- 698 question-specific 480p MP4 clips, split across ten archives.
- Parts 01–09 contain 70 questions each; part 10 contains 68 questions.
- Every archive contains:
index.html— offline annotation interface.clips/*.mp4— one video clip per question.manifest.json— question metadata without reference answers.README.md— package-level instructions.
index.jsonrecords archive sizes and SHA-256 checksums.
Reference answers are intentionally excluded from the distributed packages.
Annotation workflow
- Download one
runningbench_clips_partXX.tar.gzarchive. - Extract the archive completely. Do not open
index.htmlfrom inside an archive viewer. - Open
index.htmlin Chrome or Edge. - Enter the annotator name and review every question.
- Click 导出 JSON / Export.
- Return the generated
runningbench_partXX_annotation_<annotator>_<date>.jsonfile.
Progress is saved automatically in the browser. Each package uses a separate progress key, so multiple parts can be annotated on the same computer.
Integrity
Use the SHA-256 values in index.json to verify downloaded archives before distribution.
Source dataset
The broader RunningBench project is available at https://huggingface.co/datasets/taryya/RunningBench.