This dataset is derived from the Kinetics-400 dataset, which is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0).

Attribution

This dataset is derived from:

Original Dataset: Kinetics-400
Original Authors: Will Kay, Joao Carreira, Karen Simonyan, Brian Zhang, Chloe Hillier, Sudheendra Vijayanarasimhan, Fabio Viola, Tim Green, Trevor Back, Paul Natsev, Mustafa Suleyman, Andrew Zisserman
Original Paper: "The Kinetics Human Action Video Dataset" (https://arxiv.org/abs/1705.06950)
Original License: Creative Commons Attribution 4.0 International License (CC BY 4.0)

License

This dataset is released under the same license as the original Kinetics-400 dataset: Creative Commons Attribution 4.0 International License (CC BY 4.0)

You are free to:

Share — copy and redistribute the material in any medium or format
Adapt — remix, transform, and build upon the material for any purpose, even commercially
Under the following terms:
- Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made
- No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits

Dataset Information

Number of videos: 241181
Format: Hugging Face Dataset
Structure: Hierarchical (videos → clips → frames)
Last updated: 2025-06-04 11:55:11

Dataset Structure

The dataset contains the following fields:

video_id: Unique identifier for each video
video_path: Path to the video file
metadata: Video metadata including resolution, frame rate, quality metrics, etc.
clips: List of clips extracted from the video
- clip_name: Unique identifier for each clip
- clip_path: Path to the clip file
- start_time: Start time in the original video
- duration: Clip duration
- frames_count: Number of frames in the clip
- quality_metrics: Various quality scores
- frames: List of frames in the clip
  - frame_number: Frame number in the clip
  - image_path: Path to the frame image
  - annotation: Frame annotation
  - annotation_extra: Additional frame annotation
  - clip_score: Frame clip score
  - aesthetic_score: Frame aesthetic score

Usage

from datasets import load_dataset

# Load the dataset from Hugging Face Hub
dataset = load_dataset("liuhuanjim013/kinetics400")

# Access the data
for video in dataset['train']:
    print('Video ID: '+video['video_id'])
    print('Number of clips: '+str(len(video['clips'])))

Modifications from Original Dataset

This version of the dataset includes:

Additional frame-level annotations
Quality metrics for each clip and frame
Enhanced metadata including resolution, frame rate, and codec information
Curated subset focusing on high-quality clips

Citation

If you use this dataset, please cite both the original Kinetics paper and this derived version:

@article{DBLP:journals/corr/KayCSZHVVGNSZ17,
  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},
  title     = {The Kinetics Human Action Video Dataset},
  journal   = {CoRR},
  volume    = {abs/1705.06950},
  year      = {2017},
  url       = {http://arxiv.org/abs/1705.06950},
  archivePrefix = {arXiv},
  eprint    = {1705.06950},
  timestamp = {Mon, 13 Aug 2018 16:48:30 +0200},
  biburl    = {https://dblp.org/rec/journals/corr/KayCSZHVVGNSZ17.bib},
  bibsource = {dblp computer science bibliography, https://dblp.org}
}

Downloads last month: 92