Automated detection of animals in camera traps using deep learning in South Africa

Abstract

In the field of ecology, counting of animals to estimate population size and types of species is important for wildlife conservation. This includes analysing massive volumes of image, video or audio/acoustic data and traditional counting techniques. Automating the process of identifying, classifying, and counting animals is helpful as it phases out the tedious manual tasks of counting and labelling. The intention of this dissertation was to address manual identification and counting methods of images, by implementing an automated solution using computer vision and deep learning. This study applied a classification model to classify species, and trained an object detection model using deep convolutional neural networks, to automatically identify and determine the count of four mammal species in 1,504 images extracted from camera traps. The image classification model reported a classification accuracy of 99%, and the YOLOv8 object detection model automatically detected buffalo, elephant, rhino, and zebra school mean average precision 50 of 89%, and mean average precision 50-95 of 72.2%, and provided an accurate count over all animal classes. The results of the study showed that the application of computer vision and deep learning methods on data-scarce and data-enriched scenarios respectively, can save biologists and ecologists an enormous amount of time used on timeconsuming manual methods of analysis and counting. Additionally, these automated identification techniques can contribute towards enhancing wildlife conservation and informing future studies