paninski-lab/mirror-mouse-fused

Mirror Mouse — Lightning Pose Single-View Dataset

Single-camera pose estimation dataset for mouse locomotion keypoints, packaged for use with Lightning Pose.

Dataset Description

Head-fixed mice run on a circular treadmill while avoiding a moving obstacle (Warren et al., eLife 2021). The treadmill has a transparent floor and a mirror mounted inside at 45°, allowing a single camera to capture two roughly orthogonal views — a side view and a bottom view via the mirror — simultaneously at 250 Hz. Each frame is labeled with 14 keypoints: 7 body keypoints visible from the side (top) and the same 7 keypoints from the bottom (mirror) view.

Note: The original dataset contains keypoints on the moving obstacle; this version removes those keypoints.

Source data: Warren et al., eLife 2021 — original archive at https://doi.org/10.6084/m9.figshare.24993315. Data collected by Richard Warren and Qianyun Zhang in the Sawtell Lab, Columbia University.

Data Splits

Split	Labeled frames	Sessions
In-distribution (InD)	789	17
Out-of-distribution (OOD)	253	5

InD and OOD sets contain different sessions / animals (no overlap).

• CollectedData.csv — InD labels; videos/ — InD videos
• CollectedData_test.csv — OOD labels; videos_test/ — OOD videos

Keypoints

14 keypoints total: 7 body keypoints per view, identified by a _top (side view) or _bot (mirror/bottom view) suffix.

Top view (side)	Bottom view (mirror)
paw1LH_top	paw1LH_bot
paw2LF_top	paw2LF_bot
paw3RF_top	paw3RF_bot
paw4RH_top	paw4RH_bot
tailBase_top	tailBase_bot
tailMid_top	tailMid_bot
nose_top	nose_bot

Directory Structure

mirror-mouse-fused/
├── labeled-data/           # Extracted frames per session; includes ±2 context frames
├── videos/                 # InD session video clips
├── videos_test/            # OOD session video clips
├── videos-for-each-labeled-frame/   # 51-frame videos centered on each OOD labeled frame
├── CollectedData.csv       # InD 2D keypoint labels (x,y per keypoint)
├── CollectedData_test.csv  # OOD 2D keypoint labels
├── config_mirror-mouse-fused.yaml   # Sample Lightning Pose training config
└── project.yaml            # View and keypoint definitions (required by LP App)

The videos-for-each-labeled-frame/ directory contains 51-frame video clips with the labeled frame at the center, intended for use with temporal smoothers such as the Ensemble Kalman Smoother.

See the Lightning Pose documentation for full details on the single-view data directory structure.

Usage with Lightning Pose

The included config_mirror-mouse-fused.yaml is a ready-to-use training config. Key settings:

• Image resize: 256 × 256
• Backbone: resnet50_animal_ap10k
• Keypoints: 14
• Mirror columns: [0–6] (top view) and [7–13] (bottom view)

Update data.data_dir to absolute paths on your machine before training.

litpose train config_mirror-mouse-fused.yaml

Citation

If you use this dataset, please cite the original paper:

@article{warren2021rapid,
  title     = {A rapid whisker-based decision underlying skilled locomotion in mice},
  author    = {Warren, Richard A and Zhang, Qianyun and Hoffman, Judah R and Li, Edward Y and
               Hong, Y Kate and Bruno, Randy M and Sawtell, Nathaniel B},
  journal   = {eLife},
  volume    = {10},
  pages     = {e63596},
  year      = {2021},
  publisher = {eLife Sciences Publications, Ltd}
}

Original data archive: https://doi.org/10.6084/m9.figshare.24993315.