Description

With the increasing threat of terrorism, there is a greater need for video surveillance over a wide area. A common task of video surveillance involves the tracking of multiple targets across a camera network with overlapping/non-overlapping views. However, as the number of cameras increase, the human operator would be unable to consistently survey every video feed. Hence, this project aims to explore and develop a solution to automate the tracking of multiple targets of interest across a network of cameras with overlapping/non-overlapping views.

The primary objective of this project was to develop and implement a Small Scale Person Re-identification (ReID) system that will be deployed to test its efficiency. The accuracy of the network will be tested using various metrics such as mean Accuracy Precision (mAP) and Cumulative Matching Characteristic (CMC) curves.