Ready-made dataset

Professional AI Jogging & Locomotion Dataset — 14,500+ HQ Frames (61 videos)

Jogging-focused locomotion bundle for pose estimation, gait analysis, and motion-model training.

Engineered with the Quality Vision Motion Dataset Engine: 61 source videos, quality-first filtering, Gaussian smoothing on accepted frames, and flip + keypoint-noise augmentations.

One-time: $330Source videos: 61Rows exported (incl. augments): 14,550HQ accepted (pre-aug): 4,850

Dataset overview

  • Action label: jogging (bulk job bulk_61_videos_7efafad2)
  • Exported rows (incl. augmentations): 14,550
  • HQ frames (pre-augmentation): 4,850 accepted (~67.8% of sampled frames)
  • Mean quality (accepted): ~0.857
  • Mean landmark visibility (accepted): ~0.806
  • Stride: 3 · Post-processing: hip-centered torso scale + Gaussian smooth (window 5)

Technical stack

  • Pose + kinematics: per-frame landmarks, velocity / acceleration proxies where available
  • Layout: same export tree as other Quality Vision bulk bundles (manifest, global stats, per-video splits)

Structured layout

  • data.jsonl — merged export
  • per_video/*.jsonl, bulk_combined_manifest.json
  • features.json, global_stats.json, export_quality_report.json

Free samples

Try the public format on GitHub — QualityVision-Motion-Dataset-Samples.

Purchase & licensing

  • Price: set on the Gumroad product page (one-time purchase)
  • Deliverable: JSONL + metadata exports — not raw videos
  • Contact: info@qvision.space