CalTennis: Large Multi-View Tennis Video Dataset for 3D Human Pose

CalTennis is a large-scale video benchmark for evaluating monocular-to-3D human pose estimation in the wild. The dataset comprises over 11 million frames (51 hours) of tennis practice and match play from 40 players, captured with 2–6 synchronized cameras at 60Hz. It was created by demalenk and is hosted on Hugging Face, with a last recorded update in May 2026.

Use Cases

• Train monocular 3D human pose estimation models based on synchronized multi-view video recordings.
• Benchmark multi-view human motion capture algorithms based on expert athlete movements in tennis.
• Develop sports analytics applications based on large-scale, in-the-wild video of tennis matches and practice.

Strengths

• Over 11 million video frames, which is described as 10x larger than existing in-the-wild human motion video datasets.
• Includes 51 hours of footage from 40 different tennis players, providing a substantial subject pool.
• Features synchronized recordings from 2–6 cameras at 60Hz, enabling multi-view analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and specific file formats are unknown, which may limit suitability assessment.
• Data may reflect bias inherent to its specific collection context of tennis players and settings.

Provenance

Source

demalenk on Hugging Face

Collection Method

Captured with 2–6 synchronized cameras at 60Hz during tennis practice and match play.

Freshness

Last updated 2026-05-06 10:55:32; freshness should be verified.