MTEB logo MTEB
classification video audio

Kinetics400VA

Cite this task

@misc{kay2017kineticshumanactionvideo,
      title={The Kinetics Human Action Video Dataset},
      author={Will Kay and Joao Carreira and Karen Simonyan and Brian Zhang and Chloe Hillier and Sudheendra Vijayanarasimhan and Fabio Viola and Tim Green and Trevor Back and Paul Natsev and Mustafa Suleyman and Andrew Zisserman},
      year={2017},
      eprint={1705.06950},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/1705.06950},
}

Kinetics-400 is a large-scale action recognition dataset containing 400 human action classes from YouTube videos. Each clip is approximately 10 seconds long. Sampled 25 random shard files from each official path list, capped at 10 per class across 400 classes (~3,970 train / ~4,000 test).

Domains: ActivityWeb
Languages: English
Task type
VideoClassification
Main metric
accuracy
Source dataset
mteb/kinetics-400
License
cc-by-4.0
Dates
2017-05-19 → 2017-05-19
Annotations
human-annotated
Sample creation
found
Models scored …
Dataset preview mteb/kinetics-400 Open on HuggingFace →

Model scores

Loading…