How were the included models trained? (canny, depth, etc).

[whydna]

Anyone know if there is documentation/paper/info around the approach used to train the default models?

Looking to train my own and would be very useful. Things like data set size, diversity of data, what prompts, and other general details around general approach.

Thanks!

[geroldmeisinger]

I wrote an long article about control net training here https://civitai.com/articles/2078

[geroldmeisinger]

You can find more information about the original models here
Adding Conditional Control to Text-to-Image Diffusion Models v2 - ControlNet paper v2
Adding Conditional Control to Text-to-Image Diffusion Models v1 - original ControlNet paper
ControlNet 1.0 repo
ControlNet 1.1 repo