Asking about the training dataset

[tbaro19]

Hi there, have a nice day. I have some questions about the training dataset.
I found in provider_objaverse.py
image_path = os.path.join(uid, 'rgb', f'{vid:03d}.png')
camera_path = os.path.join(uid, 'pose', f'{vid:03d}.txt')

after I Download in https://objaverse.allenai.org/ there is only glb model , Where I can get that rgb/pose file?

[tbaro19]

Or can anyone give me some way to convert from glb, thanks !!!

[lcc815]

how do you solve it finally?

[tbaro19]

how do you solve it finally?

I used the G-Objaverse dataset =')), I think the GLB files can convert into 2 options above, but I tried and failed, can u do it ?

[apelykh]

GLB files store 3D models. They furter need to be rendered from a number of different views with given camera positions. The rendered images and camera positions are then used to train the LGM. For Objaverse, for example, there is a repo with the rendering script.