Face swap is a common computer vision application seen in many applications like Instagram, Snapchat etc. This project was to implement face swap in various scenarios, one of the scenario where face from an image is being swapped to face in the video, second is to swap two faces present in the video, itself. Third video, the face is exposed to various lightning conditions, in the dark.
Classical approach was to implement using delaunay triangulation and Thin plate splining. PRNet was used for extracting key points in deep learning approach.
Please refer report.pdf for detailed explanation and results. (Python & Tensorflow)