从真实世界失真中学习三维形状保真度指标
Learning 3D Shape Fidelity Metric from Real-world Distortions
提出基于局部连接注意力和LoRA式预训练的点云骨干网络，用于评估三维网格的感知保真度，可作为三维生成/重建任务的评价指标，替代Chamfer Distance等几何匹配指标。
原文依据 we propose a human-aligned fidelity metric that leverages local shape connectivity through a local attention mechanism
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论三维生成与重建已成为许多计算机视觉应用中的关键环节，其中重建或生成的三维形状需要在人类感知上显得真实。然而，诸如Chamfer距离之类的传统指标在比较两个三维形状时主要关注形状几何的匹配精度，无法捕捉形状的感知保真度。尽管基于频率的指标尝试在谱域分析形状细节，它们仍未能充分涵盖人类感知的复杂性。为弥补这一空白，我们提出了一种与人类对齐的保真度指标，该指标通过局部注意力机制利用局部形状连接性来捕捉丰富、详细的形状信息。我们还引入了双分支的真实形状保真度（RSF）数据集，包括一个主子集和一个仅测试子集。该数据集利用真实世界的重建与生成方法产生三维网格失真，并由数百名人类受试者进行标注。我们名为基于局部连接的形状评估（LoCaSE）的指标采用基于PointNet的骨干网络，结合低秩适配（LoRA）风格的预训练与微调来减少模型偏差，同时保持平移、旋转和尺度不变性。实验表明，与先前的指标相比，我们的方法在人类感知对齐方面取得了更优的表现。
英文原文摘要 1358 字符
3D generation and reconstruction have become essential in many computer vision applications, where the reconstructed or generated 3D shapes need to appear realistic to human perception. However, traditional metrics like Chamfer Distance to compare two 3D shapes focus primarily on matching accuracy of the shape geometry and fail to capture perceptual fidelity in the shape. While frequency-based metrics attempt to analyze shape details in the spectral domain, they still do not fully encapsulate the complexity of human perception. To address this gap, we propose a human-aligned fidelity metric that leverages local shape connectivity through a local attention mechanism to capture rich, detailed shape information. We also introduce the two-branch Real Shape Fidelity (RSF) dataset, including a main subset and test-only subset. This dataset generates 3D mesh distortions using real-world reconstruction and generation methods and annotated by hundreds of human subjects. Our metric named Local-Connection-based Shape Evaluation (LoCaSE), utilizes a PointNet-based backbone combined with Low-Rank Adaptation (LoRA)-style pretraining and finetuning to reduce model bias, while maintaining translation, rotation, and scale invariance. Experiments demonstrate that our approach achieves superior alignment with human perception compared to previous metrics.