output images are fine, but the meshes are all black about gif HOT 5 CLOSED

Did you check if you can render the meshes correctly? The model take rendered meshes as condition input. So even if you send in black images it will generate good looking images just that they will not have any condition association.

from gif.

I set the run_id=29 and after trained for a while I use the generate_random_samples.py to generate some results.
I notice that in the original dir ./outputs_files/sample/29/random_samples_q_eval_full_model, there also exist some mesh images that are black but the images are fine. So I think my train phase is normal.

I also test the generate_random_samples.py on the original trained model with run_id =29, but the saved meshes are also all black, so I think if there are some bugs in this script?

Last, I wonder how to set the flame parameters properly so that the saved mesh are normal and can see a direct corresponding semantic connection between the mesh and generated images. thanks~

from gif.

I came to notice that I can use deca to generate FLAME parameters, sorry for the bother, thanks~

from gif.

@ParthaEth, I have tried to use DECA to get FLAME parameters, but when using these parameters to generate an image, the saved mesh still all black.

Hope you can help, thanks~

from gif.

@visonpon Has this problem been solved? I have the same situation.

from gif.