Abstract: There have been significant strides in computer vision that result in momentous improvements in object detection and tracking, which form the basis of a number of applications such as surveillance, driverless vehicles and human-computer interaction. This paper proposes an original but complicated method for reliable and precise tracking based on DeepSORT (Deep Simple Online and Realtime Tracking) with YOLOv5 (You Only Look Once version 5). YOLOv5 is an effective detector that performs object detection by looking once on an image or video frame to identify objects as well as their locations. These detection results are then incorporated into the DeepSORT tracking framework, which employs deep learning techniques to consistently track objects across frames. The combination of YOLOv5 and DeepSORT addresses issues of accuracy in detecting as well as reliability in following objects thereby providing a holistic approach to dynamic scenes involving multiple objects. The proposed system detects many different yolov5s and DeepSORT at one time.
Keywords: YOLOv5, DeepSORT.
|
DOI:
10.17148/IJARCCE.2024.134116
