Wild animals often enter residential or agri-cultural areas, causing damage to property and having the potential to be dangerous to humans. This research study plan on developing a machine learning system that can automatically detect the presence of animals in surveil-lance footage and alert the appropriate authorities. Animal intrusion can lead to various problems such as including property damage, health risks, safety risks, and economic losses. Automatic animal detection has numerous applications in different fields including wildlife conservation. In this regard, two proposals utilizing the You Only Look Once (YOLO) object detection algorithm are presented. The first proposal focuses on animal detection in wildlife conservation’ reporting high accuracy in identifying and classifying animals in video frames. The second proposal addresses the issue of animal intrusion into human-inhabited ar-eas, using YOLO-based deep learning methods for object detection, tracking, segmentation, and edge detection. It also highlights the challenges of generalizing detection models from native habitats to deployment scenarios and suggests the production of synthetic data for training in a certain domain that is semi-automated. We worked on both versions YOLOv5 and YOLOv8 for detection and we received around 80% accuracy on YOLOv8. Comparatively YOLOv8 is better than YOLOv5 in terms of accuracy. These proposals showcase the effectiveness of using YOLO for animal detection and its potential to aid in animal population monitoring, behavior analysis, security, and wildlife conservation.
Animal Detection and Classification in Image & Video Frames Using YOLOv5 and YOLOv8
22.11.2023
976468 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch