Could you train a mono model?

[tumuyan]

real-esrgan-v3 has very good results and takes less time, I think if we have a mono model, the processing speed could be improved again (less data process and small file output). It is useful for text screenshots and scanned files

It would be nice to have an activation switch. user sets a color-mono threshold, and when the image is below the threshold, treats the whole image as a mono image; when the image is above the threshold, treats it as a color image

[Fischer-pixel]

i have the same trouble.

[Emanuelcarvalho2004]

Vou tentar alguma coisa mano