I have trained a WAN with gradient penalty and the loss values ​​seem to me much higher than the examples I have seen on the net. The generator receives 2 images as input and must generate a prediction of a possible middle image (such as a middle frame between two consecutive frames of a video) and the discriminator receives 3 images as input and must understand if the second image is a possible image that is between the first and third. These are the loss functions during training: enter image description here

I am training the network for another 15 epochs now and the generator loss has reached values ​​like -300 in some epochs, while in others it returns to low values. However, the results seem to be improving. Are these values ​​normal or is there something wrong?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.