Facial Recognition Technology for Identifying Cercopithecus Hybrid Monkeys

Abstract

Implementing innovative technology has become an invaluable resource for wildlife ecology and conservation studies. Recent studies support the use of facial recognition technology in primate species to assist with data collection. The purpose of this project was to develop a facial recognition software to identify individuals in a habituated population of Cercopithecus monkeys, consisting of two species and their hybrids (Cercopithecus ascanius and Cercopithecus mitis). This research is part of a long-term primate study in Gombe National Park, Tanzania. We developed an identification system by combining machine learning, object detection, and image classification. Using 16,226 images of 61 different monkey individuals, we trained an object detection system to detect the face of each monkey and we combined it with a custom-trained fast.ai Convolutional Neural Network (CNN) learning mode for identification. The sequence of these algorithms resulted in an original machine learning model with a 99.44% accuracy rate in detecting and identifying individuals from this population.