This week I focused on getting the user detection system running on the Raspberry Pi 5 with the Raspberry Pi Camera Module. I ported the vision pipeline to the Raspberry Pi and began testing live camera input while running the pose estimation model. The system uses MoveNet to detect body keypoints and estimate the user’s torso center, which is then used to compute the position error relative to the center of the camera frame. I spent most of my time testing the pipeline on the Raspberry Pi, debugging runtime issues, and verifying that the camera feed and model inference run correctly on the target hardware.
My progress is slightly behind schedule because I am still debugging a few issues related to running the system on the Raspberry Pi.
Next week, I plan to finish debugging the user detection system on the Raspberry Pi and continue testing the tracking pipeline with the camera. As I also help get the physical assembly completed, this will allow me to begin early integration testing of the full system. I will begin connecting the vision output to the motor control system so the umbrella can adjust its position based on the detected user location.
