Torment123

Results 13 comments of Torment123

Generated outputs are identical to all input images after 100,000 iterations.

Hi, I saw from the fastgan paper that in the fewshot learning datasets (100 ~ 300 images), their model is able to generate good interpolations (see Fig6 of the paper),...

question about reproducing results on sceneflow

I see, thanks for the clarification.

query about freeze BN from beginning

Thanks for your fast response.

query about freeze BN from beginning

Thanks for your fast response.

vq_gan reconstruction results blurry using default code

Thank you for the prompt detailed reply! I haven't tried the images on VQGAN yet, and am currently trying to run it. Also, about the settings in the paper, it...

vq_gan reconstruction results blurry using default code

Thanks for your clarification, this helps me a lot

vq_gan reconstruction results blurry using default code

Sure. Also, could you share some pretrained MaskGIT so that we can play with a bit? Thanks!

generalization.ckpt

I see, thanks.

Results

Hi, I'm testing the text-to-3D task with MVDream with the default settings, and the resulting 3Ds after the second refinement seem to be worse than those shown in the project...

Results

ah, never mind, I was doing text-to-3D, where the examples shown at project page is text-to-image-to3D. it was supposed to be better