This dataset contains images of primates with facial landmark annotations. Please fill out the form below and our team will review your request within 2-3 business days.

By requesting access, you agree to use this dataset for non-commercial research purposes only and to comply with all applicable institutional and ethical guidelines.

PrimateFace

Cross-species primate face analysis dataset with facial landmark annotations for non-human primates. The dataset provides annotations in two landmark conventions, enabling landmark conversion across schemes.

Landmark Conventions

PrimateFace supports two facial landmark configurations on the same images:

COCO 68-landmark (dlib68): The standard dlib/COCO Whole-Body face keypoint layout (jaw contour, eyebrows, nose, eyes, mouth).
PrimateFace 48-landmark (pf48): A custom primate-optimized landmark set.

Configs

Each config corresponds to a data source and annotation variant:

Config	Keypoints	Description
`mac-au-68kpt`	68 (dlib/COCO)	Cayo Santiago action-unit videos, COCO Whole-Body face landmarks
`mac-au-48kpt`	48 (PrimateFace)	Cayo Santiago action-unit videos, PrimateFace custom landmarks

Both configs contain the same images annotated with different landmark sets.

Usage

from datasets import load_dataset

# Load 68-keypoint annotations (dlib/COCO convention)
ds_68 = load_dataset("fparodi/PrimateFace", "mac-au-68kpt")

# Load 48-keypoint annotations (PrimateFace convention)
ds_48 = load_dataset("fparodi/PrimateFace", "mac-au-48kpt")

Citation

If you use this dataset, please cite:

@article{parodi2025primateface,
  title={PrimateFace: A Machine Learning Resource for Automated Face Analysis in Human and Non-human Primates},
  url={https://www.biorxiv.org/content/10.1101/2025.08.12.669927},
  year={2025}
}

License

CC-BY-NC-4.0 — non-commercial research use only.

Downloads last month: 8

Request access to PrimateFace