Real-Time Snake Detection And Recognition Systems In Videos Using Deep Learning

Snake species identification in videos is paramount for wildlife conservation, sustainable biodiversity, and effective snakebite management, particularly in regions prone to human-snake conflicts. Deep learning presents a promising avenue for automated recognition but faces challenges in generalizability and accuracy, particularly across diverse species and habitats. This research work proposes a novel deep learning framework for snake recognition in videos, emphasizing sustainability and human-snake conflict mitigation. Utilizing YOLO models—YOLOv6, YOLOv7, YOLOv8, and YOLOv9—for detection and segmentation, the framework is trained on a meticulously annotated video dataset sourced from snake rescuers in the Mysuru region, India. Validation results demonstrate robust performance, with mean Average Precision (mAP) exceeding 0.85 for detection and 0.94 for segmentation across various snakes. This framework offers a reliable and scalable solution for sustainable snake recognition in videos, enhancing wildlife monitoring, snakebite response, and ecological research.