Currently, the most significant risks that could jeopardize the success of our project is adapting our system to the planned demo location for the final demos. From the start of our project, we have designed our product to operate within the Hamerschlag 1300 wing. This involved calibrating camera positions to identify the ideal angle to record footage, and tuning our detection boundary and update speeds to account for the average amount of people traveling through the wing. From this, we were able to meet our use-case requirements for accuracy, prediction, and latency. However, for the final demos we will have to setup our system in the presentation room, a completely different environment that we have not tested in. Even though we are only deploying our estimation portion of the system, this new environment could have many unpredictable consequences on the performance of our product. In the coming days before the demo, we will aim to conduct tests in the final demo room, so that we can gather as much data as possible and tune our system for the new environment.
For this week, we were able to connect our database to the web application, fully integrating our system. Now, the updated estimation counts from our live footage can be seen on our web application in real time. The web application can also receive updates from multiple cameras at once, letting us synchronize our estimation results from our two main cameras. Our database is also hosted on the cloud now, so we have an unified source from which we can pull data from all of our past tests for prediction. We also continued to perform testing in the Hamerschlag 1300 wing, of which the tests and their results can be found in our testing log here. At this stage of our project, our tests were done with full system integration with 2 main cameras and 2 verification cameras. From our testing results, we realized that we needed to change the door logic after the preliminary test with the new setup that includes the use of multiple wood poles. For validation cameras, both cameras are now facing the right, so we made the necessary changes to the backend logic to accommodate for this change for validation. As a result, we were able to obtain higher percent matches between main and validation results for estimation, and also increased our prediction capabilities.
Next week, we will continue to preparing for the final demo, including testing in the new environment, and preparing our final poster and reports.