ProjFormer: Point Cloud Completion via Geometric-Projective Transformer and Cross-Modal Semantic Constraints

2026-08-15

AI Summary

This study addresses the ill-posed nature of point cloud completion and the lack of geometric consistency in existing multi-view methods by proposing ProjFormer, a cross-modal framework. By leveraging explicit 2D-3D projections to guide view attention, the method achieves deterministic geometric alignment. Furthermore, a geometry-aware routing network is designed to facilitate point-wise adaptive feature fusion and progressive optimization, overcoming the limitations of traditional fixed fusion strategies. Despite its lightweight architecture, ProjFormer delivers competitive completion performance, significantly enhancing structural integrity and geometric consistency. Consequently, this approach effectively resolves the challenge of structure recovery under sparse observations, offering a robust solution for high-fidelity point cloud reconstruction.

Abstract

Point cloud completion is inherently ill-posed due to severe sparsity and ambiguity in partial observations. Existing multi-view methods alleviate this by incorporating 2D semantics, but often rely on learned attention and fixed fusion, which lack geometric consistency and adaptability. We propose ProjFormer, a cross-modal framework that enforces geometry-consistent 2D-3D interaction through explicit projection and adaptive feature routing. A Projective Guided View Attention module aligns 3D points with multi-view features via deterministic projection, enabling efficient and geometrically consistent aggregation. Building on this, a geometry-aware routing network performs point-wise adaptive fusion of structural and observation-driven features for progressive refinement. Experiments show that, under a lightweight design, ProjFormer delivers competitive performance with improved structural completeness.

Problem

• Point Cloud Completion
• Geometric Consistency
• Cross-Modal Fusion
• Sparsity
• Ambiguity

Innovation

• Geometric-Projective Transformer
• Cross-Modal Semantic Constraints
• Projective Guided View Attention
• Geometry-Aware Routing Network
• Point Cloud Completion