TechPrufer
Sign inGet started

Blogs›Lambda›3D Object-Centric Scene Representation Learning

3D Object-Centric Scene Representation Learning

3D Object-Centric Scene Representation Learning

1
posts
2026

This post introduces a new capability in 3D scene representation learning using Deep Latent Particles (DLP). It extends DLPs to decompose real, colored 3D observations (like RGB-D images and voxel grids) into a compact set of interpretable particles, each with its own 3D position, size, and appearance. This self-supervised approach addresses challenges in 3D scene modeling, particularly the issues caused by tokenization in 2D, by focusing on object-centric representations rather than patches. The framework utilizes an appearance-aware K-means prior and a chroma loss to handle sparse voxel grids and color reconstruction. The learned representations have demonstrated effectiveness in driving robots for task completion across various benchmarks.

2026

From tokens to concepts: how particle models perceive the world

8/3/2026

Introduces 3D-DLP, a novel self-supervised approach for learning 3D object-centric scene representations. Key technical contributions include extending Deep Latent Particles (DLPs) to 3D, developing an appearance-aware K-means prior to address issues with sparse voxel grids, and implementing a chroma loss for accurate color reconstruction. The post also details experimental validation showing the superiority of 3D-DLP over baselines in manipulation tasks.