This project will investigate whether an AI-based occupancy detection system performs consistently across different demographic and environmental conditions. Using a commercial computer-vision model, the student will design and conduct an independent evaluation of person-detection performance across skin-tone groups and conditions such as lighting, occlusion, camera angle, distance, and background contrast. The project will involve dataset analysis, experimental design, statistical testing, and investigation of model failure cases. The aim is to determine whether any observed performance differences are systematic and to identify factors that may contribute to them.
The project would suit a student interested in computer vision, responsible AI, machine-learning fairness, and empirical evaluation.