This is an academic project. The main idea is to create an algorithm which recieves a static video which is crowded by people and outputs a single "clean" image with every thing else besides the people using a Detection network (YOLO V3) and classic image processing ideas to recreate the image with "clean" detected paches.
The input videos can be found in the videos page. The output images can be found in the results page.
If you are interested in learning about the work flow of this project you can read the "project description.pdf" file.