Efficient Wildlife Species Identification Using Deep Learning: A CNN-Based Approach Efficient Wildlife Species Identification

Abstract

ABSTRACT We need cheaper methods to study how animals behave in the wild, so that we can protect them and solve problems when they interact with people. Although using still and video cameras for monitoring is helpful, it often creates large amounts of data that are challenging and costly to search for specific species. In this paper, we explain a new advanced way of using computer technology to find and separate actions that are unique to certain animals. We can do this by looking at pictures or videos. We were mainly focused on telling apart animals that are closely related, like Tigers, Leopards, and Hyenas. This multiclassification job was a major challenge due of the obvious visual similarities between Leopards and Tigers. We created a deep learning model to accurately detect these species using a dataset of 2700 photos of wild animals in both zoo and wilderness settings. Our deep learning frameworks for automatic image recognition yielded remarkable results, achieving accuracy rates upwards of 98.05% for multiclassification, showcasing the effectiveness of our approach. Going beyond static images, we extended our methodology to video footage, pioneering a detection process to identify animals of interest in dynamic environments. As far as we know, this is the first time this system is being used in this situation.