ProGamerGov

@0000sir It appears somewhere in the 2x2 script, and possibly the 3x3, have an issue with the EXPAND_BORDER value that prevents the tiles from properly being put back together,

Suggestion: Multiple separate iteration values and different tiled outputs

4

The number of iterations used for the upres tiling should be an independent value of the original output's number of iterations. In practice I have found that high iteration counts...

New arxiv.org paper that looks extremely similar to this project

3

So I just found this research paper here: https://arxiv.org/ftp/arxiv/papers/1612/1612.00085.pdf, that's titled "Texture Enhancement via High-Resolution Style Transfer for Single-Image Super-Resolution". It seems to detail the same process/technique as this project....

Tiles are not correctly lined up with each other

14

Non feathered image:  Feathered image on-top of non-feathered image: 

Modalities not detected for some datasets using the Webdatasets format

2

I have found 2 examples of the modality detection code failing to recognize modalities in text and image datasets using the Webdataset format: * https://huggingface.co/datasets/ProGamerGov/synthetic-dataset-1m-dalle3-high-quality-captions * https://huggingface.co/datasets/CaptionEmporium/midjourney-niji-1m-llavanext I'm not sure...