This week I shifted my focus to work mainly on beginning to test and verify our CV backend system through different camera setups and criteria. On Friday, we permanently (for the remainder of testing time this semester) installed two cameras in the HH1300 hallway, one at the front entrance and the other in the back of the hallway next to the last small meeting room. We also mounted a new camera (with a rotatable base) on the side wall.
I was responsible for running tests for the video feed of the new camera. It turned out that the new camera provided a better field of view of the two doors that provide access to the main room. Therefore we were able to achieve better accuracy of tracking people entering and exiting the far door with respect to the camera position. Next week, my goal is to run the tests on Google Collab, which will provide better performance and stability.
Here are links to testing screenshots: New Camera Field of View, Sample Output Log. Notice that even though the resolution of this new camera seems not as good as the original cameras, the CV system was able to catch two people standing close to each other far away from the camera itself.
We will conduct a number of different tests and validations on Sunday to ensure that the Interim Demo on Monday runs smoothly. Over the next week, we will thoroughly test the backend (more information available in the Test Log) and make the necessary changes so that our tracking achieves the 80% accuracy use case requirement. That way, we will be able to shift our focus to the prediction and the frontend by the end of next week.