Foivos Paraperas

icon indicating a website link https://foivospar.github.io/ icon indicating an email [email protected]

icon company Imperial College London icon location London, United Kingdom icon location Computer Vision PhD student @ Imperial College London

Results 3 comments of Foivos Paraperas

training time

As mentioned in the supp. document of the paper, "fine-tuning an actor-specific renderer with 2 GPUs (11GB each) for 20 epochs takes an average of 17 hours".

Can the method be tested directly on images (not videos)?

Hi, Unfortunately, no. The method is designed for videos, so the Emotion Manipulator expects a sequence of length >= 10. The only implicit way to make it work for images...

About training a new neural face renderer of a new actor and the results.

Hi, If the blurred mouth appears on specific target emotions it may be that the training video is not long enough, or has not enough expression variability to properly fine-tune...