Enhancing wildlife monitoring with computer vision: A dataset for automated detection of barbary macaques

Automated monitoring of animal populations is becoming increasingly important in biodiversity research and wildlife conservation. However, detecting and identifying animals in complex and diverse environments remains a significant challenge. In this study, we introduce MagotSeg, a novel dataset designed for the automated detection, segmentation, and classification of Barbary macaques (Macaca sylvanus), a north-African primate species classified as Endangered on the IUCN red list, in the loosely environment of a zoo setting. The dataset consists of 1742 annotated images, comprising 6689 individual detections captured from multiple camera angles, at different times of day and across seasons, reflecting a wide range of environmental and lighting conditions. Unlike many existing datasets, MagotSeg includes instance-level segmentation and multi-label classification, distinguishing individuals into two age categories. This enables advanced computer vision tasks beyond basic object detection. We used the dataset to train and evaluate the performance of several versions of the YOLO (You Only Look Once) object detection model. This resource has potential applications in the automated monitoring of primates and may support conservation-related research through the integration of visual monitoring technologies.