This week, I focused on improving our multi-person tracking design. Previously, our system relied on a single speculative guessor based mainly on motion speed and past positions, which was not robust when multiple people were present. I worked on extending this approach by designing additional guessors, such as color based tracking, and integrating them under a meta selector that can dynamically choose the most reliable prediction. This design improves robustness against occlusion and ambiguity.
I also continued coordinating with my teammate on system integration planning. We developed the design review presentation slides, and I presented themin class. My progress remains on schedule, and next week I plan to begin implementing and testing the meta selection framework.