损失景观拓扑揭示了为什么简单基线在类不平衡下的3D点云分割中具有竞争力
Loss Landscape Topology Reveals Why Simple Baselines are Competitive at 3D Point Cloud Segmentation Under Class Imbalance
- KIOS Research and Innovation Center of Excellence（KIOS卓越研究与创新中心）
- University of Cyprus（塞浦路斯大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
研究3D点云分割类不平衡问题，系统评估11种方法，发现标准交叉熵表现有竞争力。通过多方面分析揭示不平衡严重程度对拓扑结构的影响，为2D有效技术难转移至3D分割提供机理解释，给出实用指导。
AI中文摘要：
3D点云的语义分割面临严重的类不平衡问题，然而2D计算机视觉中专门的不平衡感知方法在3D环境中的有效性仍不明确。我们系统地评估了11种减轻不平衡的方法，涵盖极端（641:1）和中等（56:1）不平衡率的数据集，发现标准的均匀加权交叉熵具有竞争力，通常在专门方法的0.8 - 3.3%平均交并比范围内。通过对错误模式、决策边界和优化景观几何结构的多方面机制分析，我们发现不平衡严重程度塑造了拓扑结构，极端不平衡下形成狭窄的解空间，中等不平衡下形成平坦的高原。这似乎限制了损失层面修改的有效性，所有方法都必须应对这些几何约束。我们的发现提供了实用指导，标准交叉熵提供了稳健基线，专门方法有适度改进（0.8 - 3.3%平均交并比），但调整不当可能大幅退化。这项工作首次从机制上解释了为何在2D中有效的技术不易转移到基于点的3D点云分割，并在两种代表性架构上得到验证。
英文摘要：
Semantic segmentation of 3D point clouds faces severe class imbalance, yet the effectiveness of specialized imbalance-aware methods from 2D computer vision remains unclear in 3D contexts. We systematically evaluate 11 imbalance mitigation approaches across datasets with extreme (641:1) and moderate (56:1) imbalance ratios, revealing a surprising finding: standard cross-entropy with uniform weighting achieves competitive performance, typically within 0.8-3.3% mIoU of specialized methods across architectures and datasets. Through multifaceted mechanistic analysis of error patterns, decision boundaries, and the geometry of the optimization landscape, our analyses suggest that imbalance severity shapes the topology, creating narrow solution basins under extreme imbalance and flat plateaus under moderate imbalance. This appears to constrain the effectiveness of loss-level modifications, as all methods must navigate these geometric constraints. Our findings offer practical guidance; standard cross-entropy provides a robust baseline, with specialized methods offering modest improvements (0.8-3.3% mIoU) that vary by architecture and dataset but risk substantial degradation if poorly tuned. This work provides the first mechanistic explanation for why techniques proven effective in 2D do not readily transfer to point-based 3D point cloud segmentation, validated across two representative architectures.
