Problem
Research questions and friction points this paper is trying to address.
point cloud completion
coarse-to-fine
local detail reconstruction
structural feature interaction
upsampling limitation
SUMI: Scalable Unified Model for 3D Point Cloud Inference
🤖 AI Summary
Existing point cloud completion methods struggle to recover fine local details during refinement due to simplistic upsampling and insufficient interaction with coarse structural features. This work proposes SUMI, the first approach to integrate a diffusion mechanism as a refinement module within a coarse-to-fine framework. By employing cross-attention to fuse noisy geometric features with coarse structural representations, SUMI simultaneously optimizes local details and preserves global consistency throughout the denoising process. The module is readily pluggable into existing architectures and consistently outperforms strong baselines across multiple benchmarks: it achieves state-of-the-art Chamfer Distance (CD) and F1-score on PCN, reduces CD by 16.1% on ShapeNet-55, and attains the best CD results on MVP across all output densities.
📝 Abstract
Point cloud completion commonly follows a coarse-to-fine paradigm, where a low-density coarse shape is first predicted and then upsampled to the target resolution. Although recent methods have improved global structure recovery, the fine stage often remains limited by simple upsampling and insufficient interaction with coarse structural features, making local detail reconstruction challenging. We propose SUMI, a diffusion-enhanced refinement module for coarse-to-fine point cloud completion. Unlike prior diffusion-based completion methods that use diffusion as a standalone point generator, SUMI injects noisy geometric features into cross-attention with coarse structural features, enabling reverse denoising to refine local geometry while preserving global consistency. SUMI can also be integrated into existing coarse-to-fine models as a flexible refinement module. Experiments on PCN, ShapeNet-55/34, and MVP demonstrate consistent improvements over strong baselines. SUMI achieves the best overall CD and F1-score on PCN, reduces CD by up to 16.1% on ShapeNet-55, and obtains the best CD across all output densities on MVP.
Innovation
Methods, ideas, or system contributions that make the work stand out.
diffusion-based refinement
coarse-to-fine completion
cross-attention mechanism
point cloud completion
global-local consistency
🔎 Similar Papers
No similar papers found.
