Related Experiment Video
Updated: Jul 19, 2025
16:00
Behavioral Assessment of Manual Dexterity in Non-Human Primates
Published on: November 11, 2011
22.4K
OpenMonkeyChallenge: Dataset and Benchmark Challenges for Pose Estimation of Non-human Primates.
Yuan Yao1, Praneet Bala1, Abhiraj Mohan1
1Computer Science and Engineering, University of Minnesota, Minneapolis, USA.
International Journal of Computer Vision
|August 14, 2023
Summary
We introduce the OpenMonkeyChallenge, a new benchmark for non-human primate pose estimation. This challenge utilizes a large public dataset to foster community development of generalizable pose estimation models.
Area of Science:
- Computer Vision
- Primate Behavior Analysis
- Biomedical Research
Background:
- Automatic non-human primate pose estimation is crucial for biology and biomedicine.
- Computer vision benchmark challenges have driven significant progress in related fields.
Purpose of the Study:
- To establish the OpenMonkeyChallenge, an annual competition for developing generalizable non-human primate pose estimation models.
- To create a comprehensive public dataset for training and testing these models.
Main Methods:
- Compilation of a new public dataset with 111,529 annotated images of non-human primates (17 body landmarks).
- Dataset sourced from diverse locations, including online, primate research centers, and zoos.
- Development of standardized evaluation metrics for pose estimation models.
Main Results:
- Quantitative comparison of the new dataset against existing datasets using seven state-of-the-art pose estimation models.
- Demonstration of the dataset's effectiveness in facilitating model development.
Conclusions:
- The OpenMonkeyChallenge and its associated dataset provide a robust platform for advancing non-human primate pose estimation.
- This initiative aims to drive collective community efforts towards more generalizable models.
