What can one do with an existing model?

[zackrdavis]

I'm curious about how one might use this without --is_train. I can load a model successfully, but it's not clear if there is anything in place to use it.

For instance, I would like to explore the latent space in a given region and show its linearity. Is there any simple way to input different values and output images?

[kaihuchen]

The visualize option in this code lets you sample from the trained model to generate new images. Separately I have found two others ways to make use of a trained model. First I train DCGAN on the images or videos of one specific person, then sample a straight line in the model's Z representation (which now represents the many views of this person's face) to create an animated facial expression. Another way is to train DCGAN on many faces, then use its model as a kind of reusable Universal Face Model to train on the faces of one specific person for creating this person's avatar, which results in faster training and better quality in generated images. You can see details of this in my report here: http://www.terraai.org/avatars/

[ppwwyyxx]

Another interesting to do would be vector arithmetic (as in the DCGAN paper), such as smiling woman - neutral woman + neutral man = smiling man

[georgiazhang]

@kaihuchen Hi, i'm curious about trainning videos ,can you give me some details?Thank you.

[tinacth]

@ppwwyyxx Hi! Could you share more about how to manipulate the vector arithmetic?