This week I worked on the ethics assignment and worked on the CV backend interpreting output part. I added calculating the movement of tracked people by tracking the bounding box locations of 10 past frames for each person. This is done by calculating the average center point of the person in the first 5 frames (frames 0 to 4) and the second 5 frames (frames 5 to 9), and then taking the difference between the x values of those center points, and if the difference is greater than a set offset, it will be determined as moving left or right. Originally I was calculating the difference between center points for each frame, however, after setting up our camera in the 1300 wing hallway and collecting sample footage, we found that the bounding boxes might have micro-movements across frames that would affect the results, furthermore, in some cases (occlusion, etc) bounding boxes might change significantly for certain objects, therefore I decided to use the method described above to mitigate the effects of these uncertainties. Below is a screenshot of what our image looks like after processing. I also added functionality to define doors with mouse clicks for the video feed. This is because there could be small differences between how our camera is set up during each test, and the door locations in the frame could change. Therefore it is necessary to calibrate where the doors are each time we are setting up our camera. I am currently still working on the part of counting people using the bounding boxes and movement and will finish that part tomorrow. Overall, I think I am still slightly behind schedule, but once I finish counting part tomorrow, I will be pretty much on schedule and can move on to reducing errors. Next week, I will be doing more testing to find and improve upon errors in the CV module. 