This week I mainly worked on testing the validation cameras. On Tuesday, we went to the booth workshop to make 4 wood poles that each measured at least 7 feet tall so that our cameras would not be touching the delicate walls of the HH1300 wing anymore. Once we transported the poles over to the corridor, we installed the cameras on the poles. As opposed to the main system where both cameras were installed on the same pole facing opposite directions, I employed a 1-camera-per-pole system and made the cameras both face the right hand side of the hallway. The results show that the main cameras and the validation cameras yielded videos of similar length, and while the exact accuracy of the main system have not yet reached 80%, we can confirm that there is a high match between the main and validation outputs, which is definitely some good progress.
The web app and prediction components are still in the works because much of it cannot be completed until we collect enough verified data, but we are on track to complete our use-case requirements before the final video is due. Over the next week, we will continue to perform the remaining tests in the lab, and I am planning on to finish the prediction part by next Sunday. We will do more validation tests to ensure the 80% accuracy use-case requirement is fulfilled.
