VIP-200K-Video: 200,000 Video Clips from YouTube

HiDream-ai provides a collection of video files from the VIP-200K dataset, packaged as 100 tar archives. The repository contains only the raw video clips, with annotations and metadata available separately. The dataset was last updated on March 19, 2026.

Use Cases

• Train video-based AI models based on the raw video clips.
• Benchmark video processing pipelines using the structured tar archive format.
• Develop computer vision applications leveraging the YouTube-sourced video content mentioned in the description.
• Create multimodal datasets by combining these videos with the separate annotation files.

Strengths

• Contains video files from a dataset referenced as having 200,000 instances.
• Organized into 100 tar archives for structured batch access.
• Clips are organized by YouTube video ID, suggesting a consistent source.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• The repository contains only video files, requiring a separate download for annotations and metadata.

Provenance

Source

HiDream-ai

Collection Method

Likely sourced from YouTube, based on the description of video clips organized by YouTube video ID.

Freshness

Last updated 2026-03-19 10:11:48; freshness should be verified.