I want to generate my image

Thanks for your wonderful work. I'm trying to use your method and pretrained model to generate relighting images. The albedo image seems correct, but I find the predicted values of the front normals['full'] are all 0 with different input latent codes. Did I use the model in a wrong way, hope for your respones, Thank you very much.

Hi!

Thanks for your awesome work of enabling HDRI lighting manipulation in the facial NeRF! The design is really inspiring.

I observe that the lighting manipulation relies on the augmented dataset, which is generated by the Total Relighting according to the paper.

Since the Total Relighting is not open-source, I wonder whether it is possible to release your augmented dataset for boosting relevant studies?

Thanks in advance!

Hi, awesome work!

I have tried the notebook, but I wonder how to fix the pose while rotating the environment map?

I have tested:

• Horizontally or Vertically shifting the HDRI map as arrays.
• Rotating the HDRI map as arrays.
• Explicitly passing the camera_orientation to the infer_relit.

However, none of them give me the results shown in the project page which features the obvious moving shadow.

Thus, I wonder what's the correct way to rotate the environment map? Thanks!

Hi, thanks for your awesome work!

I wonder how to get the preconvolved environments map given a HDR environments? Thanks!