Closed EricZgw closed 5 years ago
Codes about this repo along with some pretrained models will be released around the end of May. The relevant inpainting code is not fully suitable for this outpainting task. The context normalization and relative spatial variant loss proposed in our paper are practical modules for this task.
I try to just change the location of the mask, using the inpainting method. But the effect is not good. Can you give me some suggestions?