This project aims to estimate the pose of non-human primates (monkeys) as part of the OpenMonkeyChallenge. This project was taken up as the final project for CSCI 5561: Computer Vision coursework at the University of Minnesota under the guidance of Dr. Junaed Sattar.
OpenMonkeyChallenge (OMC) is an ongoing computer vision benchmark challenge for NHP (Non-Human Primate) pose tracking.
The data for this challenge has been made publicly available by Yao et. al.. The data has been collected from various sources including the internet, National Primate Centers, and the Minnesota Zoo which contains a diverse species (26 species) of NHPs. It is a set of 111,529 images annotated with 17 landmarks - Nose, Left eye, Right eye, Head, Neck, Left shoulder, Left elbow, Left wrist, Right shoulder, Right elbow, Right wrist, Hip, Left knee, Left ankle, Right knee, Right ankle, and Tail.
Our project uses a combined architecture of rCNN and CPMs to classify the species and estimate their pose.