search

nv-tlabs / GET3D

Other
4.2k stars 376 forks source link

A bit confused on how this works #75

Closed mihailpettas closed 1 year ago

mihailpettas commented 1 year ago

So what exactly is this model? Can i input any type of image? Or do i need to train it on the objects i want?

SteveJunGao commented 1 year ago

Hi @stojanni ,

This is a 3D generative model, which samples a random latent code (from a Gaussian Distribution), then generates a 3D mesh with textures from this latent code. Therefore, we don't need to input any type of image.

To do this, we do need to first train on the object that we want the model to generate (e.g. chairs or motorbikes)

SteveJunGao commented 1 year ago

Close this issue as no update received