Final Year Project - Target Tracking Using Camera For Unmanned System

This FYP is done in AY 2021/2022.

Summary: The end goal is to formulate a Real-Time Multiple Object Tracking (MOT) Algorithm to be used in an embedded device called the DJI Manifold 2G, onboard a drone. However, due to challenges faced when integrating software with proprietary hardware (Manifold 2G), the hardware is changed to Nvidia Jetson AGX Xavier to continue with the project. The Algorithm chosen was YOLOv5+DeepSORT.

• YOLOv5 training information can be seen here.
• AGX installed with: Python3.8, Pytorch 1.7.0, Torchvision 0.8.1. (Mainly referred to this link)
• Evaluation done using TrackEval.