kingsting
kingsting
There is a statement for choose negative examples “neg_keep = npr.choice(len(neg), size=len(neg), replace=True)” in gen_imglist_pnet.py. The augment "replace=True" will lead to repetitive examples, so why not use "replace=False" to get...
I tried fastdtw, very slow . Anyone has this problem?
Do you run this model on GPU or CPU? If runs on CPU, how could it archive real time performance with so large network?
The link https://github.com/joshua-wu/deepfakes_faceswap/blob/master/anonfile.com/p7w3m0d5be/face-swap.zip is unavailable.