This week, I finally got the Jetson Nano working and did its initial setup. I also have detection for all five ingredients running, shown below. Tofu is outlined in magenta, fish balls in cyan, beef in red, bok choy in green, and mushrooms in blue. Using an HSV color picker tool, I significantly narrowed down the ranges for the beef, bok choy, and mushrooms to cut out most of the false hand detections. Since this was done on footage from a cell phone, the ranges might need adjusting when we apply the algorithm to footage from the IMX camera.
There are still some incorrect categorizations but they are infrequent enough to be covered by our majority rule strainer classification plan. The only errors that may be the exception to that are the tofu vs. fish balls, which frequently get crossed. I will be trying to fix that a bit once we get our final setup footage with the Jetson Nano and the IMX camera this coming week. Getting the final setup footage will be essential to putting in the last tweaks in the algorithm and also for determining which strainer ingredients are dropped into. With only a month left in the course, I am feeling a little bit rushed to start getting things integrated. However, I think that we are overall in a good place to get our MVP done in time.