Tetris3D: 3D Scene Generation With Objects That Fit Together
What it is
Tetris3D is a generative framework for single-image 3D scene reconstruction that ensures objects are physically and geometrically coherent. It works by explicitly conditioning the generation of each object on the geometry and physical relationships of its neighbors, and it introduces ComOb, a physics simulation-based dataset of 1.2M scenes with per-object meshes and pairwise relation annotations.
The problem it solves
Existing 3D scene reconstruction methods generate objects independently or couple them implicitly, failing to ensure fine-grained spatial and physical compatibility between interacting neighboring objects.
What you could use it for
- Reconstructing 3D scenes from single images where objects interact physically
- Generating 3D object shapes and poses that remain stable and plausible in complex scenes
- Training generative models on the ComOb dataset of physically interacting object scenes
Read the source ↗Back to the library →
Summary written by an AI model from the source; check the source before you rely on it.