search

CompVis / latent-diffusion

High-Resolution Image Synthesis with Latent Diffusion Models
MIT License
11.48k stars 1.5k forks source link

Training algorithm for inpainting from scratch #197

Open morteza89 opened 1 year ago

morteza89 commented 1 year ago

Could you please help stepwise on how to train the algorithm on our dataset? Thanks

BenjaminIrwin commented 1 year ago

I'm also interested in this. @morteza89 have you had any luck?

morteza89 commented 1 year ago

Not really. I need to train in on my custom dataset, did all the data preprocessing suggested to do based on the LAMA algorithm but still not lucky to start training...

BenjaminIrwin commented 1 year ago

Ok. We're in a very similar position. We will contact you and potentially can collab on this, if you're interested.

morteza89 commented 1 year ago

Ok. We're in a very similar position. We will contact you and potentially can collab on this, if you're interested.

Sounds great.

DMAgit commented 1 year ago

Any update on this?

ustczhouyu commented 1 year ago

Hello, I would like to ask the difference between unconditional LDM and conditional LDM. After the model is trained, is unconditional sampling generate image randomly, but not based on a given image? So, if I want to generate a normal image from a flawed image (without any annotations in the inference phase), should I use conditional LDM? @morteza89 @BenjaminIrwin @asanakoy @pesser

ksai2324 commented 1 year ago

Any updates on this? I also want to train my own inpainting model, but it seems like there is no documentation available on how to do that