problems about predicting normal and depth independently
snowflakewang opened this issue · comments
Hello, thank you for your brilliant work in Stable Diffusion-based geometry estimation!
I wonder whether your model can predict normal and depth independently. i.e. I can input an image and the model will only output a depth or normal.
Thank you :)
Hi, thanks for your interest! However, one of our contributions is to utilize mutual guidance of depth & normal to improve each other (see ablation study). Thus it must generate the depth & normal maps that share the same spatial layout at the same time. Sorry for it.
thank you for your answer!