2 ms·
>they're too unstable', a widespread myth >See for example BigGAN I remember when you try training a BigGAN model on anime images, the quality was bad. Now lo
by GaggiX 3y ago
>they're too unstable', a widespread myth
>See for example BigGAN
I remember when you try training a BigGAN model on anime images, the quality was bad. Now look at this example, one single GPU, 1.5M images with a diffusion model: https://medium.com/@enryu9000/anifusion-diffusion-models-for-anime-pictures-138cf1af2cbe https://medium.com/@enryu9000/anifusion-diffusion-models-for... ,the difference in quality is absurd, you can say this or that is not true but the quality speak for itself, obtaining good quality on complex distribution is much easier with a diffusion model than a GAN.
For example in the case of the site linked they have conditioned the model on poses because you're not going to get anything close to be coherent without them with a simple StyleGAN as they say they're using.
- Tyr42 3y agoBroken link?
- GaggiX 3y agoFixed thx
- gwern 3y ago> I remember when you try training a BigGAN model on anime images, the quality was bad Because there was a bug in the code, in a part unrelated to the GAN itself. > the difference in quality is absurd Yes, it does help to train on anime with code that isn't buggy. (BTW, Skylion was getting good results with GANs on anime similarly restricted to centered figures like those samples, he just refuses to ever publish anything.)
- GaggiX 3y agoSo you believe that without the bug you would be able to come close to the quality of the diffusion model I have linked? I'm not even asking about using the same compute (1 GPU for ~1 month) but if you just believe BigGAN can come close to that in general. Also the bug is probably related to the added complexity of training a GAN model in comparison to a diffusion model.