search

cvlab-kaist / 3DGAN-Inversion

Official Implementation of WACV 2023 paper "3D GAN Inversion with Pose Optimization".
MIT License
110 stars 13 forks source link

3d shape and depth map extraction #6

Closed juanjomon closed 1 year ago

juanjomon commented 1 year ago

First, congrats for your interesting work, but I have a question:

How can I get the 3d shape in mrc file from a given photo that has been inverted? So far the output is only a mp4 video. How can I get the correspondent 3d shape? (as the ones generated with gen_samples from the ffhq pkl).

On the other hand.. is it possible to obtain the depth map video as the one you display in your project website aside the image video?.

Any help would be much appreciated.

Thanks.

mlnyang commented 1 year ago

Hi thinks for your comment!

We haven't try on generating video from mrc file, but this issue may help you. -> https://github.com/NVlabs/eg3d/issues/66

For the depth map video, you can just utilize the video generation code https://github.com/KU-CVLAB/3DGAN-Inversion/blob/3cfebf9abc0733aae5c5e512f33ce18d016e3e48/training/coaches/single_id_coach.py#L62 and change the image_mode = 'image_depth' in here

juanjomon commented 1 year ago

thanks.