search

thuanz123 / realfill

Unofficial implementation of RealFill
MIT License
359 stars 27 forks source link

the data generated by outpainting has obvious discontinuous traces at the mask boundary. #10

Closed Lucy-IM closed 9 months ago

Lucy-IM commented 12 months ago

Hello author, your work is really great. When studying your work, I found that using the parameters and test data provided by the author, the data generated by extrapolation has obvious discontinuous traces at the mask boundary. Is there any way to eliminate it.

I really hope to receive your letter.

Lucy-IM commented 12 months ago

0

thuanz123 commented 11 months ago

First of all, I have to say that I'm not related to the author in any way and this is an unofficial implementation.

Secondly, yeah it is because the way I blend the original masked image with the output of the model is very naive so sometimes it will show a small white boundary

Lucy-IM commented 11 months ago

Thank you very much for receiving your answer！ @thuanz123.

Lanxin1011 commented 10 months ago

Thank you so much for providing this implementation @thuanz123!!!. The fundamental reason for the small white boundary might be the irregular value of the mask, the same as I've encountered in an inpainting work. I checked the value of the mask image in '/data' and found the mask values have other values apart from 0 and 255 at the boundary of the masked region, which may result in blurry edges or the above white boundary.

thuanz123 commented 9 months ago

Thks @Lanxin1011!!! After using correct mask, the issue seems gone so I think I can safely close this issue now. Simply correcting the mask and the masked image does the trick here