During this week, my main focus was on continuing the integration between the our web streaming program running on the Raspberry Pi, and the computer vision backend running on Google Colab. Our CV backend is now capable of accessing multiple camera streams hosted on multiple ngrok servers concurrently. Due to limitations on the free version of ngrok, each of these servers had to be hosted on separate ngrok accounts, associated with unique API keys that are divided across the Raspberry Pis we are using. No complex system was required to synchronize our video feeds. Each Raspberry Pi independently pushes their video stream to their own ngrok server, from which a child thread in our CV backend program continuously polls for the most recent frame posted to the server.
My portion of the project remains on schedule. My upcoming objective is to continue running tests to tune the accuracy of our estimation system. Now that our camera streaming and computer vision systems are fully integrated, we can fully focus on gathering as much test data to process and make adjustments to our detection model off of. Thanks to the independent nature of our web streaming subsystem, we can even split up the CV processing of our main detection and secondary verification cameras across multiple Google Colab instances, greatly increasing our computational capacity. A goal for next week is to record more test sessions with multiple cameras on Google Colab, while also finishing our Final Presentation slides.