Towards Implementation of Detection and Tracking of Wild Animals
R.I. Viji, G. Sreelatha, Sandeep S. Kumar · 2023
The conflicts between human and animal in regions closer to forest areas leading to destruction of resources, loss of human life and domestic animals etc. have become a subject of great concern. Advanced algorithms in object detection and visual object tracking can assist in developing proper monitoring systems that can help in reducing the causalities by providing warnings while preserving animal life. A tracking-by-detection approach is presented in this paper, detection of animals done by Yolov4 which is trained using custom image dataset for 9 specific classes. The trained model is used as descriptor in DeepSORT tracker to track the movement of animals along video frames. The detector has an mAP accuracy of 98% and the performance evaluation of the tracker is done using the benchmark dataset AnimalTrack.