# Spectrum AUC Difference (SAUCD): Human-aligned 3D Shape Evaluation

arXiv: https://arxiv.org/abs/2403.01619 · HTML: https://mlsanity.com/paper/2403.01619/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 34 of its 83 references that are in the map.
- First posted on arXiv 3 Mar 2024.

## Builds on (most cited first, 30 of 34)

- [PyTorch: An Imperative Style, High-Performance Deep Learning Library](https://mlsanity.com/paper/1912.01703/) (2019)
- [Pixel2Mesh: Generating 3D Mesh Models from Single RGB Images](https://mlsanity.com/paper/1804.01654/) (2018)
- [Convolutional Mesh Regression for Single-Image Human Shape Reconstruction](https://mlsanity.com/paper/1905.03244/) (2019)
- [What Do Single-view 3D Reconstruction Networks Learn?](https://mlsanity.com/paper/1905.03678/) (2019)
- [Fast Differentiable Sorting and Ranking](https://mlsanity.com/paper/2002.08871/) (2020)
- [3D Point Cloud Generative Adversarial Network Based on Tree Structured Graph Convolutions](https://mlsanity.com/paper/1905.06292/) (2019)
- [Total3DUnderstanding: Joint Layout, Object Pose and Mesh Reconstruction for Indoor Scenes from a Single Image](https://mlsanity.com/paper/2002.12212/) (2020)
- [FaceVerse: a Fine-grained and Detail-controllable 3D Face Morphable Model from a Hybrid Dataset](https://mlsanity.com/paper/2203.14057/) (2022)
- [Enriching Local and Global Contexts for Temporal Action Localization](https://mlsanity.com/paper/2107.12960/) (2021)
- [Multimodal Shape Completion via Conditional Generative Adversarial Networks](https://mlsanity.com/paper/2003.07717/) (2020)
- [Two-Stream Consensus Network for Weakly-Supervised Temporal Action Localization](https://mlsanity.com/paper/2010.11594/) (2020)
- [DeepHandMesh: A Weakly-supervised Deep Encoder-Decoder Framework for High-fidelity Hand Mesh Modeling](https://mlsanity.com/paper/2008.08213/) (2020)
- [DECOR-GAN: 3D Shape Detailization by Conditional Refinement](https://mlsanity.com/paper/2012.09159/) (2020)
- [Learning to Generate 3D Shapes from a Single Example](https://mlsanity.com/paper/2208.02946/) (2022)
- [Min-Max Similarity: A Contrastive Semi-Supervised Deep Learning Network for Surgical Tools Segmentation](https://mlsanity.com/paper/2203.15177/) (2022)
- [PC-HMR: Pose Calibration for 3D Human Mesh Recovery from 2D Images/Videos](https://mlsanity.com/paper/2103.09009/) (2021)
- [Learning Delaunay Surface Elements for Mesh Reconstruction](https://mlsanity.com/paper/2012.01203/) (2020)
- [Towards Generic Image Manipulation Detection with Weakly-Supervised Self-Consistency Learning](https://mlsanity.com/paper/2309.01246/) (2023)
- [Point Scene Understanding via Disentangled Instance Mesh Reconstruction](https://mlsanity.com/paper/2203.16832/) (2022)
- [Cycle-guided Denoising Diffusion Probability Model for 3D Cross-modality MRI Synthesis](https://mlsanity.com/paper/2305.00042/) (2023)
- [High Fidelity 3D Hand Shape Reconstruction via Scalable Graph Frequency Decomposition](https://mlsanity.com/paper/2307.05541/) (2023)
- [Language-guided Human Motion Synthesis with Atomic Actions](https://mlsanity.com/paper/2308.09611/) (2023)
- [Deep Hybrid Self-Prior for Full 3D Mesh Generation](https://mlsanity.com/paper/2108.08017/) (2021)
- [Learning to Generate and Reconstruct 3D Meshes with only 2D Supervision](https://mlsanity.com/paper/1807.09259/) (2018)
- [Task-Oriented Low-Dose CT Image Denoising](https://mlsanity.com/paper/2103.13557/) (2021)
- [Fostering Generalization in Single-view 3D Reconstruction by Learning a Hierarchy of Local and Global Shape Priors](https://mlsanity.com/paper/2104.00476/) (2021)
- [Self-supervised Human Mesh Recovery with Cross-Representation Alignment](https://mlsanity.com/paper/2209.04596/) (2022)
- [Progressive Multi-view Human Mesh Recovery with Self-Supervision](https://mlsanity.com/paper/2212.05223/) (2022)
- [MeshMVS: Multi-View Stereo Guided Mesh Reconstruction](https://mlsanity.com/paper/2010.08682/) (2020)
- [Reducing Positional Variance in Cross-sectional Abdominal CT Slices with Deep Conditional Generative Models](https://mlsanity.com/paper/2209.14467/) (2022)

## Built on by (most cited first, 9 of 9)

- [MH-pFLID: Model Heterogeneous personalized Federated Learning via Injection and Distillation for Medical Data Analysis](https://mlsanity.com/paper/2405.06822/) (2024)
- [IDOL: Unified Dual-Modal Latent Diffusion for Human-Centric Joint Video-Depth Generation](https://mlsanity.com/paper/2407.10937/) (2024)
- [AnatoMask: Enhancing Medical Image Segmentation with Reconstruction-guided Self-masking](https://mlsanity.com/paper/2407.06468/) (2024)
- [dFLMoE: Decentralized Federated Learning via Mixture of Experts for Medical Data Analysis](https://mlsanity.com/paper/2503.10412/) (2025)
- [Disease-informed Adaptation of Vision-Language Models](https://mlsanity.com/paper/2405.15728/) (2024)
- [FSC: Few-point Shape Completion](https://mlsanity.com/paper/2403.07359/) (2024)
- [Divide and Fuse: Body Part Mesh Recovery from Partially Visible Human Images](https://mlsanity.com/paper/2407.09694/) (2024)
- [SRAM: Shape-Realism Alignment Metric for No Reference 3D Shape Evaluation](https://mlsanity.com/paper/2512.01373/) (2025)
- [Textured Geometry Evaluation: Perceptual 3D Textured Shape Metric via 3D Latent-Geometry Network](https://mlsanity.com/paper/2512.01380/) (2025)

## Abstract (arXiv)

Existing 3D mesh shape evaluation metrics mainly focus on the overall shape but are usually less sensitive to local details. This makes them inconsistent with human evaluation, as human perception cares about both overall and detailed shape. In this paper, we propose an analytic metric named Spectrum Area Under the Curve Difference (SAUCD) that demonstrates better consistency with human evaluation. To compare the difference between two shapes, we first transform the 3D mesh to the spectrum domain using the discrete Laplace-Beltrami operator and Fourier transform. Then, we calculate the Area Under the Curve (AUC) difference between the two spectrums, so that each frequency band that captures either the overall or detailed shape is equitably considered. Taking human sensitivity across frequency bands into account, we further extend our metric by learning suitable weights for each frequency band which better aligns with human perception. To measure the performance of SAUCD, we build a 3D mesh evaluation dataset called Shape Grading, along with manual annotations from more than 800 subjects. By measuring the correlation between our metric and human evaluation, we demonstrate that SAUCD is well aligned with human evaluation, and outperforms previous 3D mesh metrics.