Deep learning for person re-identification
Amena Khatun · Queensland University of Technology · 2021
This thesis addresses the problem of correctly re-identifying a target person in a crowded environment, in a multi-camera surveillance system to ensure the safety of people in mass gatherings. Using deep neural networks, we provide effective solutions to the challenges caused by variations in lighting conditions, viewing angles, background, and occlusion in a camera network, and demonstrate the efficacy of the novel algorithms and frameworks that we have developed for accurate person re-identification in real world scenarios.