I think that my social distance detector was kind of effective. I believe it was able to identify the number of social distance violations well, but it could definitely be more effective. Some improvement could come with increasing from 720p to 1080p quality so that the better-quality video enables the program to more accurately analyze distance between people. Also, a more accurate detector could come from having a camera angle that is more vertical or higher above the people walking.
I do not think this approach would be very accurate in estimating new infections, because there are too many individual variables when it comes to determining the vulnerability a person has to getting COVID-19. Two 25-year-olds in good health walking five feet away from each other poses a different infection threat than say a 25-year-old who is walking at least 6 feet away from everyone else but sneezes without covering their mouth, so the sneeze still reaches about ten feet (they have strong lungs). Then that is just concerning the individuals who could infect others, not the people who could become infected. If an individual with a compromised immune system is six feet away from the person who sneezed, but they are still within the sneeze zone, then technically it would never be identified by the detector as a possible infection. Yet, due to the individual’s personal medical history they could become infected. In today’s circumstance I believe the variables that should be taken into account are distance, time, previous medical history, and precautions (masks, washing hands, hand sanitizer, etc.). Obviously, it is very illegal for a program to be able to facially recognize someone then access their medical history to determine their vulnerability. However, with cellphone data organizations and state governments have already been able to track people’s movements and the amount of time they stayed in that location. This also brings up a lot of ethical questions concerning privacy vs public safety.
A limitation is the legality and ethical dilemma about whether video should even be recorded of people. An improvement could be making the process as transparent as possible with instead of the distance violation being tracked and notified away from the actually source/people who committing the violation and might not even know that they are being watched and are violating anything, they are notified of their possible violation.