about step1 vqgan

[supgy]

Hi author, When I trained vqgan in step1 with my own dataset, the generated samples are as follows, they seem to have a lot of noise at the edges, and some generated samples cannot distinguish between foreground and background. Could you give me some advice about this? (my dataset:465 CTs with size 128128128 and I did not modify the configuration in synt_ct.yaml and vqgan3d.yaml)

[qic999]

@supgy Thanks for your attention to our work! There might be some issues with the results saved in the log. You can perform the inference again and manually save the results to review them.