@joshua-wu
thanks for your wonderful work. I have some questions that hope you can help me:
this code is about swap trump&cage, if i want to swap another people's face to trump or cage, do i need to train from start or fine tune on the trained model or just to change the input?
i notice the side face and when face was covered by hand or other things, the results is not good, do you think adding those data will improve the results?
how about have high resolution image as training data? will it improve results ignore the compute complex?
@joshua-wu thanks for your wonderful work. I have some questions that hope you can help me:
this code is about swap trump&cage, if i want to swap another people's face to trump or cage, do i need to train from start or fine tune on the trained model or just to change the input?
i notice the side face and when face was covered by hand or other things, the results is not good, do you think adding those data will improve the results?
how about have high resolution image as training data? will it improve results ignore the compute complex?