In Order To Detect And Track Wildlife In Real Time, Computer Vision Techniques Are Being Used More And More In Wildlife Monitoring. Modern YOLO Object Detectors (Yolov3, Yolov4, Yolov5, Yolov7, And Yolov8) Combined With Multiobject Tracking Algorithms, Specifically SORT And Deepsort, Are Assessed And Contrasted In This Study For Automated Wildlife Monitoring. Wildlife Camera Trap Datasets Are Used To Evaluate These Models\\\' Performance, Taking Into Account Metrics Like Tracking Accuracy, Precision, Recall, Mean Average Precision (Map), And Inference Speed.According To Experimental Results, Deepsort Considerably Increases Tracking Stability By Lowering Identity Switches Through Appearance-Based Association, While Yolov8 Achieves The Best Detection Performance In Terms Of Map And AP@0.5. When Paired With Deepsort, Yolov5 Offers A Robust, Lightweight Baseline That Achieves High Tracking Accuracy (MOTA ≈ 94%) While Utilizing Computational Power Efficiently. Conversely, SORT, Which Has More Identity Switches And Only Uses Motion Cues. The Results Show The Trade-Offs Among YOLO Variants In Terms Of Detection Accuracy, Model Size, And Computational Cost. The Suggested YOLO + Deepsort Framework Shows Great Promise For Real-Time Wildlife Monitoring On Edge Devices Like Uavs And Field Cameras, Supporting Applications Like Habitat Analysis, Biodiversity Assessment, Antipoaching Surveillance, And Mitigating Conflicts Between Humans And Wildlife.
M et al. (Thu,) studied this question.