CenterSnap: Single-Shot Multi-Object 3D Shape Reconstruction and Categorical 6D Pose and Size Estimation

TL;DR

CenterSnap introduces a real-time, single-shot method for multi-object 3D shape reconstruction and 6D pose estimation from RGB-D images, outperforming existing multi-stage approaches especially in complex scenarios.

Contribution

The paper proposes a novel one-stage, bounding-box free approach that jointly predicts 3D shape, pose, and size for multiple objects in a single forward pass.

Findings

Achieves real-time processing at 40 FPS.

Outperforms existing methods with a 12.6% absolute improvement in 6D pose accuracy.

Effectively handles occlusions and complex multi-object scenes.

Abstract

This paper studies the complex task of simultaneous multi-object 3D reconstruction, 6D pose and size estimation from a single-view RGB-D observation. In contrast to instance-level pose estimation, we focus on a more challenging problem where CAD models are not available at inference time. Existing approaches mainly follow a complex multi-stage pipeline which first localizes and detects each object instance in the image and then regresses to either their 3D meshes or 6D poses. These approaches suffer from high-computational cost and low performance in complex multi-object scenarios, where occlusions can be present. Hence, we present a simple one-stage approach to predict both the 3D shape and estimate the 6D pose and size jointly in a bounding-box free manner. In particular, our method treats object instances as spatial centers where each center denotes the complete shape of an object along with its 6D pose and size. Through this per-pixel representation, our approach can reconstruct in real-time (40 FPS) multiple novel object instances and predict their 6D pose and sizes in a single-forward pass. Through extensive experiments, we demonstrate that our approach significantly outperforms all shape completion and categorical 6D pose and size estimation baselines on multi-object ShapeNet and NOCS datasets respectively with a 12.6% absolute improvement in mAP for 6D pose for novel real-world object instances.