Great work! Runs quickly and very well
The first person to implement depth2img into spaces! Love this, keep up the excellent work!
yes I've build the Gradio demos Gradio for the official SD release so It was easy to make it run here.
We're implementing the pipeline on Diffusers and it will be much much easier to use
Would it be possible to use this to generate the depth map from the image?
Is a depth map generated when it is not provided? If so could that be output as well?
Currently the way diffusers pipeline is implemented, the generated depth image is not returned https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_depth2img.py#L537-L598
You could modify this script to enable that. The depth image is generated under the hood by a DPT model like this https://huggingface.co/spaces/nielsr/dpt-depth-estimation