Open Mark-Dou opened 3 years ago
Just extract all faces and not swap faces from the same person.
Thanks for replying me, but a new problem appears that frames in the same videos always select frames in another same videos to blend cause the fixed threshold in find_one_neighbor.How do you solve this problem?
The original paper said blending data samples are composited using real images in FF++,but each video we extracted lots of frames, how do we select the sourceImages?each video select one image or something?I would appreciate it if you could solve my problems?Thank U