-
Notifications
You must be signed in to change notification settings - Fork 26
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Proper Image Got into Faulty Mosaic Effect after Upscaled #63
Comments
I get the same result using |
Hi @cherrynoize , sorry for the buggy powershell scripting. As a matter of fact, I am working on it but in a separate project repository (since it's rare to find a tested posixshell+powershell standard library). Appreciate your patience. FYI, next release would be the last one using Shell script wrapping method as I will personally go into the C & C++ codes to make everything scratch. This can be caused by 2 issues:
|
Hm, that is really odd behaviour. I´ll look into it and try to fix it :) |
That would be great. BTW, I thought I had answered to @hollowaykeanho mentioning I wasn't on Windows so I didn't think any |
So you arent running this on windows but on linux? If yes, then this is not an issue for me, but for @hollowaykeanho |
I am having the same issue with the squares being random in the image. It does not happen on every run, but it seems to happen a lot. I have tried about 6 models and they all seem to do the same thing, so it does not seem to be a model issue. Image poisoning is somewhat impossible as these are textures from a 20 year old game. Could this simply be some issue in the image save function? The squares in the image look ok, just not in the right location in the over-all picture. |
Wow. This is definitely the first case I've seen. @cherrynoize, and @uq1, can both of you send me a the original image to be included in the test suite please? I will analyze it. @Joly0, no worries. Please work on the parallel execution task ya. =) I will triage the case. Apologize for this the late responsiveness. Am currently involved in a client's project. Will patch the software in and out. |
FYI: I just upscaled the same image again and this time it worked fine. The original was being called through a game, while this time I tried it just from command prompt. Could other processes CPU/GPU usage be affecting it? I hope this insight helps. |
@uq1 , not an issue here. It should lower level like drivers or hardware. I'm incline to former. |
Yeah... same case for the Ubuntu (I'm on Debian). It should be related to either model or low level driver before we arrive to hardware (@uq1 , your case is very unlikely a hardware problem). Ubuntu is known for heavy modifications for better user experience. |
Help wanted to further root cause the problem. In the meantime, I do what I can to provide workaround for this weird bug. |
I'm seeing the same issue but only with certain partially AI-generated video. I've tried many of the models, and for certain input sources it always makes a mosaic. The generated AI images are fine looking - I had hoped to improve them 2x, but when I try to upscale them they come out mosaic'ed. |
@hollowaykeanho For me this happens with videos and only on |
Really appreciate someone putting some effort into a CLI version for this. I would love to work on some PRs myself but I just don't have the time right now.
So at least I would like to point out that there's some issues with saving the output image, as by default with param
--input image.jpg
I would get an error for it tries to save toimage.jpg/image-upscaled.jpg
(which is clearly not what's expected). Specifying--output
per se does not solve the issue but I found that--input ./image.jpg --output ./image-upscaled.jpg
(and probably omitting output too) works around the issue. Though it does seem to ignore the--output
param completely.More importantly though:
I get this.
Input is:
I would love for this to work properly so I could implement it in my dotfiles to auto upscale wallpapers so I could keep them in a more reasonable format until needed. Maybe it's the model? I'll try different ones now.
Also sorry for making this all a single issue. Maybe I should have made two.
The text was updated successfully, but these errors were encountered: