[TRAINING] stable diffusion/ lama

[bogeyturn]

I am considering training models for inpainting and would ask if someone is willing to help collect data for the dataset & help actually train it. There are multiple inpainter I want to try out.
Without hint:

• lama
• stable diffusion

With hint:

• Stable Diffusion with ControlNet

Dataset:

• Danbooru2021
• clean dataset?!
• maybe other datasets too?
• use existing manga & use detector
• use pertained models(stable diffusion)

[Wenaka2004]

my inpainting ControlNet may may meet your needs
https://huggingface.co/Wenaka/NoobAI_XL_Inpainting_ControlNet_Full

[Schwi-AI]

Approximately how many training samples are needed? And what kind of annotation is required?
I actually considered continuing to train the DeepCreamPy model to specialize in the painting styles of certain artists.
Also tried some inpainting model with ComfyUI. Those models work very well at body editing, but they fell short in terms of refining the close up NSFW details.