Description
Pool Table Video Dataset — 210TB (AI/ML Training + Vision R&D)
A 210-terabyte high-volume video dataset focused on pool/billiards tables, captured to support computer vision and AI/ML training use cases. The archive prioritizes consistent framing, clear visibility of table geometry and ball motion, and real-world variation (lighting, angles, felt colors, environments, camera motion), making it well-suited for tracking, detection, and physics-aware video modeling.
What’s inside (high level)
-
Extensive filmed coverage of pool tables and gameplay scenarios
-
Variation across camera angles, table styles, lighting conditions, and background environments
-
Long-form footage and/or clip-ready segments (depending on how you want it packaged)
-
Organized dataset delivery with consistent naming + folder structure
Great for
-
Ball detection + multi-object tracking (multi-ball tracking, occlusions, collisions)
-
Table geometry understanding (pockets, rails, perspective, calibration)
-
Event detection (break shot, pocketed ball, fouls, turns)
-
Physics-aware modeling (trajectory prediction, contact events, spin/deflection research)
-
Robotics / automation (shot planning, perception for cue sports systems)
Optional add-ons (recommended for higher-value buyers)
-
Clip curation into training-friendly segments (e.g., 5–30s / 30–120s clips)
-
Metadata package (CSV/JSON): camera angle, lighting, table type, felt color, scene tags
-
Annotation services (choose one or combine):
-
Ball bounding boxes + IDs (tracking)
-
Pocket/rail/table keypoints (geometry)
-
Event labels (shot start/end, pocketed, collision moments)
-
-
Compliance-focused filtering (face/license plate blurring if any bystanders appear)
Delivery
Delivered as a structured dataset via cloud transfer (for subsets) and/or physical drives for full-volume delivery. Dataset can be shipped as raw archive, curated clips, or pre-split train/val/test packages.
Dataset name: Pool Table Video Dataset
Total size: 210TB
Format: Video (source and/or export formats available)
Content: Pool tables + gameplay-focused footage
Primary ML tasks: Detection, tracking, geometry, event detection, trajectory prediction
Packaging options: Raw archive / curated clips / train-val-test splits
Add-ons: Metadata CSV/JSON, annotations, compliance filtering


Reviews
There are no reviews yet.