Timothy Sainburg at Harvard Dataverse produced this collection of 3D skeletal keypoints for approximately 350 mice across 8 species and 18 subspecies. The data consists of 30-minute open-field recording sessions captured at 120-150fps using a synchronized 6-8 camera array.
Use Cases
- Classifying behavioral motifs using the MoSeq syllables column
- Analyzing locomotive dynamics through speed and turn velocity metrics
- Developing 3D pose estimation algorithms using the triangulated keypoint positions
Strengths
- High-speed capture at 120-150fps
- Taxonomic diversity covering 18 subspecies
- Includes derived behavioral MoSeq syllables
Limitations
- Restricted to open field environment behavior
- Potential for triangulation noise in 3D keypoint inference
Provenance
- Source
- Timothy Sainburg, Harvard Dataverse
- Collection Method
- Triangulation of 2D keypoints from a synchronized 6-8 camera array
- Freshness
- Last updated 2026-03-10