LionSense: Redefining Wildlife Monitoring and AI-Powered YOLOv5 Lion Detection and Classification

The use of artificial intelligence (AI) in wildlife conservation has become increasingly significant during the past few years. This has lead the way for responsible actions and creative strategies in a number of disciplines relating to wildlife, including species discovery, habitat monitoring, and conservation initiatives. This study mainly focuses on the classification and detection of lions using the YOLOv5 algorithm and bounding boxes. The research leverages computer vision, image classification, and deep convolutional neural networks to achieve accurate results. The EfficientNet_b3 pre-trained model weights are utilized to overwrite the default weights in YOLOv5. The researchers manually labeled and annotated the lion images for training the models. They built individual classification models for various lion features such as eyes, ears, nose, whiskers, and the full face. The accuracy of the classification models varied, with the full face and whiskers models achieving the highest accuracy. The models were loaded and integrated into a pipeline that considers both the full face and whiskers models to improve overall accuracy. The lion classification model achieved an accuracy of 86.8%, while the lion detection model achieved 75.44% accuracy. Overall, this research demonstrates the effectiveness of using YOLOv5 and deep learning techniques for unique lion classification. The study highlights the importance of considering specific lion features and building individual models to achieve the highest accuracy. The proposed pipeline integrating the full face and whiskers models provides promising results for lion classification and detection tasks.