This week I mainly worked on improving the prediction ML algorithm. At the start of the week the accuracy of the algorithm was stuck between 0.35 to 0.40 given the model data. After a series of tweaks and changes to the program, we were able to achieve a prediction accuracy of 0.73-0.80 by the end of the week with the model data. Once the system is integrated and real data can be fed from the backend, we are hopeful that we will be able to showcase this feature for the interim demo on April 3 and achieve the 0.80 prediction accuracy we set in our design review.
In terms of the web app, there is no significant progress to report since last week. We have been focusing on getting the system integrated between the camera hardware and the backend, and that will be our priority up to the interim demo. So, according to the Gantt chart schedule we set, the progress of the web application is now behind. However, given that the Django project is already setup, I will be able to link the prediction and current occupancy to the web app once integration is done between the camera and the backend.
In the coming week, as the interim demo is coming up, the backend remains a top priority. If we are able to connect everything to the backend and test its outputs before the weekend, I am hopeful that I will be able to at least connect the estimation feature to the front end before the interim demo on April 3.