LATTE-MV: Multiview Sports Video with Human Pose and Ball Tracking

A multiview sports video dataset released by ember-lab-berkeley on 2025.04.14. The dataset likely contains clipped match videos alongside synchronized tracking data for balls, paddles, and human keypoints in 2D and 3D. The data structure suggests it is intended for computer vision tasks in sports analytics.

Use Cases

• Train human pose estimation models based on the 2D and 3D keypoints files.
• Develop ball and paddle tracking algorithms based on the provided CSV tracking results.
• Conduct multiview action analysis in sports using the synchronized video and metadata.
• Benchmark 3D human pose reconstruction from multiple camera views.

Strengths

• Provides synchronized multimodal data including video, CSV tracking files, and numpy keypoint arrays.
• Includes both 2D and 3D human pose keypoints, which suggests depth information is available.
• Dataset structure is organized by match and clip, which may facilitate systematic analysis.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, video duration, and dataset scale are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

ember-lab-berkeley via Hugging Face

Collection Method

Likely recorded and processed from sports matches; method details are not specified.

Time Range

null

Freshness

Last updated 2025-04-14 02:17:17; freshness should be verified.

Geography

null