Unsupervised Point Cloud Completion through Unbalanced Optimal Transport

2024-10-03

arXiv.org

AI Summary

Real-world point cloud completion faces challenges due to the scarcity of complete ground-truth annotations and the difficulty of acquiring paired training data. Method: This paper proposes the first unpaired point cloud completion framework based on Unbalanced Optimal Transport (UOT). By formulating the mapping from incomplete to complete point clouds as an UOT problem, it eliminates reliance on rigid alignment assumptions and paired supervision, thereby enhancing robustness and generalization under severe class imbalance. The approach integrates a lightweight point cloud feature encoder with a differentiable UOT solver for end-to-end optimization. Contribution/Results: The method achieves state-of-the-art (SOTA) or SOTA-comparable performance on both single-class and multi-class standard benchmarks. The source code is publicly available.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Computer Vision: Learning & Optimization for CV

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Security and Privacy: Data transparency and provenance

Abstract

Unpaired point cloud completion is crucial for real-world applications, where ground-truth data for complete point clouds are often unavailable. By learning a completion map from unpaired incomplete and complete point cloud data, this task avoids the reliance on paired datasets. In this paper, we propose the extit{Unbalanced Optimal Transport Map for Unpaired Point Cloud Completion ( extbf{UOT-UPC})} model, which formulates the unpaired completion task as the (Unbalanced) Optimal Transport (OT) problem. Our method employs a Neural OT model learning the UOT map using neural networks. Our model is the first attempt to leverage UOT for unpaired point cloud completion, achieving competitive or superior performance on both single-category and multi-category benchmarks. In particular, our approach is especially robust under the class imbalance problem, which is frequently encountered in real-world unpaired point cloud completion scenarios. The code is available at https://github.com/LEETK99/UOT-UPC.

Problem

• Unpaired point cloud completion without ground-truth data
• Formulating completion as Unbalanced Optimal Transport problem
• Robust performance under class imbalance scenarios

Innovation

• Uses Unbalanced Optimal Transport (UOT)
• Neural networks learn UOT map
• Robust to class imbalance