Published September 3, 2026 | Version v1

AquaPose YH Tutorial Dataset

Authors/Creators

  • 1. Georgia Institute of Technology

Description

A 30-second, 12-camera tutorial clip for AquaPose (3D fish pose estimation via refractive multi-view triangulation). Contains trimmed H.264-encoded video from all 12 cameras, refractive calibration JSON, trained YOLO detection and pose estimation weights, a platform-neutral pipeline config.yaml, and pre-computed reference outputs. Intended to let an outside researcher reproduce the AquaPose pipeline end-to-end on real data. LICENSING: This deposit is mixed-license. The videos (videos/) and refractive calibration (geometry/calibration.json) are released under CC-BY-4.0. The bundled model weights (models/yolo_obb.pt, models/yolo_pose.pt) are AGPL-3.0-derived artifacts, having been trained with Ultralytics (AGPL-3.0), and are NOT covered by the CC-BY-4.0 grant. AquaPose itself is AGPL-3.0-or-later. Cite the DOI when using the videos or calibration data.

Notes

Contents extract to a directory named 'aquapose-tutorial-data/'. Verify the download with 'cd aquapose-tutorial-data && sha256sum -c checksums.sha256' before use; all 22 files should report OK.

Files

aquapose-tutorial-data.zip

Files (215.6 MB)

Name Size Download all
md5:dd53c066091d772a446043e4c6792b33
215.6 MB Preview Download