Multiple Object Tracking using Yolov7 Based Object Detection and Norfair Tracking

Abstract

The computer vision techniques have advanced video processing and intelligence generation for a variety of difficult dynamic circumstances. Tracking moving objects is incredibly useful in many computer vision applications. This work provides a method for tracking multiple objects in a video.  It entails, first detecting the object using YOLOv7 model and then giving detection results to Norfair framework. Norfair is a customizable, lightweight Python framework for monitoring 2D objects in real time. The Manhattan distance function is employed to assess the distance between recently observed items and the objects it is already monitoring. From the result it has been analyzed that the proposed model considerably enhances tracking performance while handling occlusion, appearance changes and fast moving object.