GigaPixel vs another?

LOL, going back to the original other (are we allowed to name it) had a play with it.

Very limited models, fails when you try to use another scale than the one the model was built for,
this could maybe worked around by a better front end.

Scaling one 4320 × 2160 image took ~2.5 hours compared to a few seconds in TopazLab’s tools. although I had to use CPU AI as some of the pytorch features are not supported on my hardware yet.

The Denoising only model didn’t work as well as Denoising and SR combined model, very inferior to DenoiseAI.

Be nice if some of the paper it was based on’s idea could be incorporated by TopazLabs though as it did seem to SR text a bit better than TopazLab’s tools.

1 Like

yes, agree with everything you told. but on some specific picture, sometimes, it did better than gigapixel on tiny picture with face for example. for everything else, it’s useless. but it’s always interesting when we don’t achieve an acceptable result with our software at topaz to see what the others can do. my issue is that the online version is limited to 640x480 picture, after that we get a cuda out of memory issue.
Tried to install it locally and can’t make it working so far. can i write to you , if you installed it locally , to see where i did wrong ? thx !

Yes I installed it locally, running on my M1 Mini Mac, happy to try and help but if the error is cuda related I probably won’t be much more help than a google search :slight_smile:

1 Like