halfjoe / 3d-portrait-stylization Goto Github PK

Hello, thanks for offering the source code of your wonderful work!
When I try to run your main_recon_deform.py in Pycharm on win10 with GPU, I find that the PhotoAvatarLib.exe outputs nothing. So I can not get the xxxx_face_fit_ortho.obj for following code. Why it happens?

Hi, Thanks for open-sourcing this project. Is it possible to run on a Ubuntu instead of Windows? To do so, what needs to be changed? Thanks

I open the file "3D-Portrait-Stylization-main\data_demo_results\00016\result_texture_transfer\00016_face_fit_ortho_final.obj" by Blender.

However, it shows an undesired result as follows.

I would like to ask why, or what other software is used to open it。

Thanks a lot!

The Landmark Translation code (train and test) is the core of the paper. I hope that the code is opened soon. Thanks very much!

First, the ratio of aspect is not correct.
Second, a black area exists in the boundary of image.
Third, the quality of stylization is not perfect as in your demo.

After successfully running the Face Reconstruction & deformation I'm encountering the following error when running the Texture Stylization step (transfer/main_texture_transfer.py) on the provided test data (00016):

Traceback (most recent call last):
  File "./transfer/main_texture_transfer.py", line 199, in <module>
    run(dir_data_sub)
  File "./transfer/main_texture_transfer.py", line 127, in run
    images = renderer_model(azi=azi[aov_id], ele=ele[aov_id])
  File "/home/gpu/anaconda3/envs/3d-portrait-stylization/lib/python3.7/site-packages/torch/nn/modules/module.py", line 547, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/gpu/avatarTech/3D-Portrait-Stylization/texture_style_transfer/transfer/model_ortho_RE.py", line 61, in forward
    elevation=self.elevation, azimuth=self.azimuth, show_forehead=True)
  File "/home/gpu/avatarTech/3D-Portrait-Stylization/texture_style_transfer/neural_renderer/renderer.py", line 185, in render_rgb
    textures = torch.cat((textures, textures.permute((0, 1, 4, 3, 2, 5))), dim=1)
RuntimeError: CUDA out of memory. Tried to allocate 1.36 GiB (GPU 0; 15.90 GiB total capacity; 14.49 GiB already allocated; 869.44 MiB free; 64.90 MiB cached)

What are the GPU requirements to run transfer/main_texture_transfer.py?