Shape-Prior-Based Point Cloud Completion for Single-Stage Fully Sparse 3D Object Detection

2026-05-30

AI Summary

Sparse and incomplete point clouds severely degrade the performance of one-stage fully sparse 3D object detectors in autonomous driving. To address this limitation, this work proposes a proposal-free point cloud completion method that first identifies foreground points directly via an instance selection module and then performs structured completion using shape prior prototypes aligned by object centers and orientations. This approach effectively recovers missing geometric details, leading to substantial improvements in detection accuracy. Evaluated on the KITTI dataset, the proposed method consistently enhances the performance of two representative one-stage fully sparse detectors, demonstrating its effectiveness and generalizability.

Abstract

Single-stage fully sparse 3D object detectors rely on point clouds data to detect objects in autonomous driving scenarios. However, the sparsity and incompleteness of point clouds significantly limit the performance of 3D object detection. To address this issue, this paper proposes a point clouds completion method specifically designed for single-stage fully sparse detectors. The entire shape-prior-based completion process consists of two consecutive steps. In the first step, we design a novel Instance Selection module, which is capable of identifying point clouds corresponding to foreground objects even when the baseline model does not generate proposals, while effectively ignoring the point clouds of background regions. In the second step, we introduce a novel Alignment-Based Point Completion module, which aligns the point clouds of foreground objects with prototypes in terms of both their centers and orientations. Subsequently, points are selected from the prototype to fill in the missing parts of the foreground object. We evaluated our method on two single-stage fully sparse detectors using the KITTI dataset. The experimental results demonstrate that the proposed method significantly improves the detection performance, confirming its effectiveness and generalizability.

Problem

• point cloud completion
• 3D object detection
• sparsity
• incompleteness
• single-stage fully sparse detectors

Innovation

• shape prior
• point cloud completion
• single-stage 3D detection
• instance selection
• alignment-based completion