This week I helped with integrating the camera system with the treadmill and verifying that QR codes could be scanned. We tested the camera at varying heights and positions on the treadmill and marked the region that we found was the best for positioning (far enough from the head of the treadmill to place boxes and high enough to see the entire width of the treadmill).
Since we have completed a lot of the kinematics software already, I wanted to add an ML component to the project, so I decided to start working on a system that detects objects on a treadmill. My idea was that instead of QR codes labelling the category of each box, which can be tedious in real life, we could have the algorithm determine the object using CV. This is a reach goal for our project because getting an accurate classier for a variety of objects and integrating it into the system is a difficult task. Nevertheless, this week I started coding up the classification algorithm using PyTorch and Torchvision.
