Does GET3D have image-guided shape generation feature?

Technical Blogs & Events Connect With Experts - AMA GET3D Team- AMA May 4th, 2023
,
1

Hey there!
in your press release (https://developer.nvidia.com/blog/rapidly-generate-3d-assets-for-virtual-worlds-with-generative-ai/) there is a phrase: “GET3D is a new generative AI model that generates 3D shapes with topology, rich geometric details, and textures from data like 2D images, text, and numbers.” - so, it seems GET3D can somehow generate a 3D model from a picture… I mean something like that: at the Inference stage, GET3D takes a png image as a parameter, and in response generates not a bunch of random 3D models, but a 3D model similar to png picture. Is it possible to use GET3D this way, if yes - how exactly (it would be great to get any docs or advices)…
Vasil.

2

Hi, Thanks for the questions! Due to time constraints, we are not working on image-guided shape generation for GET3D, we might explore that in the future.

1 Like
3

thanks for the answer!
is there any understanding when there will be a version with the ability to upload an image?

4

We do not have a specific timeline for it but will be actively looking into this

5

From the point of view of technology, is it not possible to upload a picture and get a 3D model from it now? or is there such an option, but not available to us (developers)?

6

Hi, we do not have this option at the moment, and we would encourage developers/researchers to try PTI inversion if having one image as input and wants to generate 3D shape from the image GitHub - danielroich/PTI: Official Implementation for "Pivotal Tuning for Latent-based editing of Real Images" (ACM TOG 2022) https://arxiv.org/abs/2106.05744

7

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.