https://youtu.be/3L3GrMPIk0k
Yes, the social distance detector was effective at detecting potential violations. The boxes around the individuals would change colors dependent on whether or not they were close to each other. There are some issues when an individual gets too close or too far from the camera however. It is harder to determine if they are closer or further away. The detector seems to be using the Check() function and uses the center of the individuals to determine if people are getting too close together. It returns True if people are too close and False if they are keeping their distance.
The speed of the program is the only thing which concerns me about the effectiveness of estimating for new infections. On my system, it takes over a third of a second to study each frame, which means to study a full 30 frames (1 second) it will take over 10 seconds. While there could be some modifications to improve the speed, the machine may also lose some of its accuracy. This system could be used at shops and stores to confirm that people are maintaining social distance protocols. Through the security cameras one could determine if the individual businesses are correctly following social distance policies.
Some limitations to the proposed design are the invasion of privacy. While this system is not neccessarily intrusive, it does bring into question surveillance of citizens. Improvements to the design could be improving the speed of the calculations, or to run the system with faster machines. Also, the accuracy at a distance and up close to the camera is somewhat questionable, so it would be important to improve these areas.