Closed maliho0803 closed 4 years ago
The dataloader loads all the data for the embedder, generator and discrimninator. In the training loop each network receives its proper input. e.g. x_hat = G(g_y, e_hat)
@vincent-thevenin @maliho0803 In the original paper, section 3.2, it is said the input to the generator is not used for the embedder
But it seems that, in your dataloader, the input to the embedder is K images and K landmarks, while the input landmark to the generator is randomly picked from the K samples.
as original paper saied 'The generator G(yi(t); e^i; ; P) takes the landmark image yi(t) for the video frame not seen by the embedder,the predicted video embedding e^i and outputs a synthesized video frame x^i(t). ', however, in your dataloader part, the image used in generator was appeared in embedder?