SPAN: Spatial-Projection Alignment for Monocular 3D Object Detection

TL;DR

This paper introduces SPAN, a novel method for monocular 3D object detection that enforces geometric consistency through spatial and projection alignment, significantly improving detection accuracy.

Contribution

The paper proposes Spatial-Projection Alignment (SPAN) with spatial point and 3D-2D projection alignment, addressing geometric inconsistency in decoupled 3D attribute prediction.

Findings

Improves monocular 3D detection accuracy

Integrates seamlessly with existing detectors

Demonstrates significant performance gains

Abstract

Existing monocular 3D detectors typically tame the pronounced nonlinear regression of 3D bounding box through decoupled prediction paradigm, which employs multiple branches to estimate geometric center, depth, dimensions, and rotation angle separately. Although this decoupling strategy simplifies the learning process, it inherently ignores the geometric collaborative constraints between different attributes, resulting in the lack of geometric consistency prior, thereby leading to suboptimal performance. To address this issue, we propose novel Spatial-Projection Alignment (SPAN) with two pivotal components: (i). Spatial Point Alignment enforces an explicit global spatial constraint between the predicted and ground-truth 3D bounding boxes, thereby rectifying spatial drift caused by decoupled attribute regression. (ii). 3D-2D Projection Alignment ensures that the projected 3D box is aligned tightly within its corresponding 2D detection bounding box on the image plane, mitigating projection misalignment overlooked in previous works. To ensure training stability, we further introduce a Hierarchical Task Learning strategy that progressively incorporates spatial-projection alignment as 3D attribute predictions refine, preventing early stage error propagation across attributes. Extensive experiments demonstrate that the proposed method can be easily integrated into any established monocular 3D detector and delivers significant performance improvements.