The most significant risk that would jeopardize the success of our project is whether or not we will be able to meet the accuracy requirements for tracking and prediction. During our initial testing, we found that there have been issues with the tracking algorithm when handling cases of occlusion. This is likely due to the internal constraints of object detection and tracking algorithms. Specifically, during our initial testing, if there is a person walking in front of the camera for a period of time, blocking the field of view and obstructing the person behind in the hallway, there is no way for the object detection algorithm to identify the person behind, and object tracking algorithm will run into ID switching errors due to losing track of the object. Furthermore, if the person walking in front of the camera blocks the entrances to the individual rooms, it is almost impossible to detect people entering/exiting the individual rooms accurately. Therefore it might be necessary to rescope the requirements about the number of people in the hallway at once to handle these cases of occlusion, as given our current one-camera setup for monitoring the hallway, it would be difficult to handle these cases of occlusion without significant system design changes. We will do more testing of our system in the upcoming week and decide if a redefinition of our requirements would be necessary.
During this week, we have not made any changes to the existing design of the system.
We have slightly changed our schedule in that Gary is now working on the backend prediction module instead of working web application during the past week(after discussion with the professor and TA, we decided it is more important to finish developing/testing our backend first). In the upcoming week, we will perform further testing on our whole camera module and backend system (getting video feed, counting, and prediction) as a team. Below are images showing the testing we did last week using our camera footage and CV module:
