KungfuAthleteBot: Martial Arts Training Motion Capture Data

197 video clips from athletes' daily martial arts training, processed into 1,726 sub-clips via automatic temporal segmentation. The data was processed using GVHMR for motion capture and GMR-based reorientation. It was authored by LuluCao and last updated on May 8, 2026.

Use Cases

• Train human motion synthesis models based on processed motion capture data.
• Develop action recognition systems for martial arts movements.
• Benchmark temporal segmentation algorithms on athletic video data.
• Analyze motion continuity and transitions in physical training sequences.

Strengths

• Contains 197 original video clips from real-world martial arts training.
• Processing resulted in 1,726 sub-clips designed to minimize abrupt motion discontinuities.
• Motion data has been processed through GVHMR capture and GMR reorientation pipelines.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the processed motion data is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Athletes' daily martial arts training videos.

Collection Method

Automatic temporal segmentation, GVHMR motion capture, GMR-based reorientation.

Freshness

Last updated 2026-05-08 08:20:37; freshness should be verified.