Ready-made dataset

High-Quality Running Pose Dataset

Train your AI faster with clean, ready-to-use motion datasets.
No preprocessing. No noise. Just results.

This is a clean, processed Running Motion Pose Dataset containing over 4,500 high-quality frames extracted from 20 different running videos.

One-time: $250Action: runningFormat: JSONL

Key Features

  • Total Frames: 4,500+ HQ frames (after strict quality filtering + augmentations)
  • Processing: Gaussian temporal smoothing (window=5), hip-center body normalization, horizontal flip + keypoint noise augmentation
  • Pose Format: Full 33 keypoints (COCO-style + hands & feet) with visibility scores
  • Additional Data: Sequence features, per-frame quality scores, motion consistency metrics, and full quality reports

Quality Metrics

  • Mean Quality Score: 0.86
  • Mean Landmark Visibility: 0.79
  • Mean Motion Consistency: 0.52 – 0.64 (typical for running movements)

Perfect for

  • Training pose estimation models
  • Human motion analysis
  • Action recognition (running vs walking)
  • Sports AI and biomechanics
  • Imitation learning and animation

Structured data layout

  • data.jsonl — global merged dataset (all processed frames)
  • per_video/*.jsonl — per-source splits
  • features.json — sequence-level motion features
  • low_quality_frames.jsonl — rejected frames for hard-example mining
  • manifest.json, global_stats.json, export_quality_report.json — metadata and QA

Free samples

Validate the format on GitHub: QualityVision-Motion-Dataset-Samples.

License

Proprietary / Research & Commercial use allowed (no redistribution of the raw dataset).

Note: This is a pure Running Pack. A Mixed (Walking + Running) Pack and a dedicated Walking Pack are also available.

Purchase & licensing

  • Checkout: Gumroad (one-time $250; see product page)
  • Deliverable: JSONL + metadata exports — not raw source videos
  • Contact: info@qvision.space
Buy $250 on Gumroad Contact for custom datasetsinfo@qvision.spaceBack to pricing