Facemap — Lightning Pose Single-View Dataset

Single-camera orofacial pose estimation dataset for mice, packaged for use with Lightning Pose.

Dataset Description

Mice were head-fixed and recorded with a close-up camera of the face during neural activity experiments. 15 keypoints are labeled around the eyes, nose, mouth, whiskers, and lower lip; a paw keypoint is also included for frames in which the paw enters the camera field of view. Several keypoints (e.g., paw) are frequently occluded or absent.

Sessions were recorded from two camera angles (cam0 and cam1), both treated as a single view.

Source data: original archive at https://doi.org/10.25378/janelia.23712957.

Data Splits

Split	Labeled frames	Sessions
In-distribution (InD)	2,400	57
Out-of-distribution (OOD)	100	6

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

Keypoints

15 keypoints total.

Region	Keypoints
Eye	eye_back, eye_bottom, eye_front, eye_top
Nose	nose_bottom, nose_r, nose_tip, nose_top, nosebridge
Mouth	lowerlip, mouth
Whiskers	whisker_c1, whisker_c2, whisker_d1
Paw	paw

Directory Structure

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

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

Usage with Lightning Pose

The included config_facemap.yaml is a ready-to-use training config. Key settings:

• Image resize: 256 × 256
• Backbone: resnet50_animal_ap10k
• Keypoints: 15

Update data.data_dir to an absolute path on your machine before training.

litpose train config_facemap.yaml

Citation

If you use this dataset, please cite:

@article{syeda2024facemap,
  title     = {Facemap: a framework for modeling neural activity based on orofacial tracking},
  author    = {Syeda, Atika and Zhong, Lin and Tung, Renee and Long, Will and
               Pachitariu, Marius and Stringer, Carsen},
  journal   = {Nature Neuroscience},
  volume    = {27},
  number    = {1},
  pages     = {187--195},
  year      = {2024},
  publisher = {Nature Publishing Group US New York}
}

Original data archive: https://doi.org/10.25378/janelia.23712957.