Sabitlenmiş Tweet
Excited to share our recent work! 😁
An Object is Worth 64x64 Pixels:
Generating 3D Objects via Image Diffusion
We generate meshes as a set of UV patches packed in an image, which we term as "Object Images" or "omages”.
Code is now live! omages.github.io
8 cool facts about omages:
1. Omage encodes geometry as a (R,R,3+1) image, which is essentially a 16-bit RGBA PNG!
2. PBR material is encoded in another (R,R,8) image.
So three PNGs will give you a realistic 3D object.
3. You can use image generation models to generate 3D objects, one image for one object!
4. Discrete patch structures emerge out of continuous noise during the denoising process.
5. Patch segmentation comes naturally from 2D disconnected components, no instance labels needed!
6. Irregular connectivity & complex topology?
No worries, all encoded in a regular image.
7. Change your object resolution by just rescaling the omage!
8. Generated shapes come with UV maps—no unwrapping required.
X. And there’s even more to discover!
#omages
English
