NAM-YOLOV7: An Improved YOLOv7 Based on Attention Model for Animal Death Detection

Dead animals on the road are very harmful to public health. Because of offensive odours and the potential outspread of diseases, dead animals endanger public health. Existing methods focused on collisions between vehicles and animals on roads, number of animals, protection of crops from animals etc. To solve this issue, two main tasks listed below can be used: (1) Detecting dead animals on the highway and (2) Notification to the appropriate authorities. In this paper, we explore the viability of object detection methods to detect dead animals. We scrutinize and compare various versions of the "YOLO"(you only look once) in detecting dead animals. We compare the performance with the improved YOLOv7 model with the earlier versions when trained on the ADD (Animal Death detection) dataset and results show that improved YOLOv7 performs best when compared to the earlier YOLO models.