Closed gulinsong closed 3 years ago
Hi @gulinsong , they are randomly sampled when testing; We have not added the option to obtain a shape / appearance code for a given input image. This could be done with test-time optimization of the latent codes, or adding an encoder during training.
Hi , thanks for your great work. I want to know about how to get shape and appearance codes, are these encoded from an image when training, and generate randomly when testing?