← all work generative-vision-lab

Realesrgan

Result

The finding

Notes

- Trained with L1 + adversarial. - Fine-tuned from the published x4 checkpoint into an x6 architecture, with the upsampling stage rebuilt for a factor of 6. - 20,000 steps on 720 training frames, evaluated on held-out val and test splits.

Output

Same subject across the row; each column is the same stage in every project.

Elsewhere

Code and full results on GitHub

Every number is read from the file the run produced. github.com/hammasbuilds · kaggle.com/muhammadhammas13