Abstract
The advancement in object tracking involves the integration of feature-based approaches with contemporary deep learning methodologies. The primary difficulties in object tracking pertain to the establishment of reliable data associations across consecutive frames. These challenges are particularly pronounced in scenarios involving surveillance and autonomous navigation. The you only look once version 5 small (YOLOv5s) detector trained on the VisDrone2019 dataset results in a notable reduction in latency. The proposed methodology demonstrates superior performance compared to baseline approach, with F1 score of 93.31 for Intersection over Union (IOU) values greater than 0.5, while achieving a frame rate of 167.147 frames per second within a mere 0.0024 seconds. Experimental results are presented.