mlsanityData to 14 Oct 2026

Robustifying Point Cloud Networks by Refocusing

arXiv 2308.05525 ↗First posted 10 Aug 2023cs.CV cs.LGcited by 2

Not on the map. Why?

Abstract (arXiv)

The ability to cope with out-of-distribution (OOD) corruptions and adversarial attacks is crucial in real-world safety-demanding applications. In this study, we develop a general mechanism to increase neural network robustness based on focus analysis. Recent studies have revealed the phenomenon of textitOverfocusing, which leads to a performance drop. When the network is primarily influenced by small input regions, it becomes less robust and prone to misclassify under noise and corruptions. However, quantifying overfocusing is still vague and lacks clear definitions. Here, we provide a mathematical definition of focus, overfocusing and underfocusing. The notions are general, but in this study, we specifically investigate the case of 3D point clouds. We observe that corrupted sets result in a biased focus distribution compared to the clean training set. We show that as focus distribution deviates from the one learned in the training phase - classification performance deteriorates. We thus propose a parameter-free refocusing algorithm that aims to unify all corruptions under the same distribution. We validate our findings on a 3D zero-shot classification task, achieving SOTA in robust 3D classification on ModelNet-C dataset, and in adversarial defense against Shape-Invariant attack. Code is available in: https://github.com/yossilevii100/refocusing.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Builds on 53 of its 67 references that are in the map. First posted on arXiv 10 Aug 2023.

Builds on

53 of 67 references.

References · 53

  1. 2016Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localizationcited by 6,733
  2. 2016PointNet: Deep Learning on Point Sets for 3D Classification and Segmentationcited by 5,248
  3. 2017PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Spacecited by 4,401
  4. 2018Dynamic Graph CNN for Learning on Point Cloudscited by 2,223
  5. 2020Shortcut Learning in Deep Neural Networkscited by 1,969
  6. 2019Natural Adversarial Examplescited by 1,408
  7. 2017Ensemble Adversarial Training: Attacks and Defensescited by 840
  8. 2020PCT: Point cloud transformercited by 732
  9. 2016Multi-View 3D Object Detection Network for Autonomous Drivingcited by 724
  10. 2021Point-BERT: Pre-training 3D Point Cloud Transformers with Masked Point Modelingcited by 653
  11. 2020Transformer Interpretability Beyond Attention Visualizationcited by 627
  12. 2022Rethinking Network Design and Local Geometry in Point Cloud: A Simple Residual MLP Frameworkcited by 379
38 more
  1. 2019On the Relationship between Self-Attention and Convolutional Layerscited by 300
  2. 2019Analyzing the Structure of Attention in a Transformer Language Modelcited by 243
  3. 2020PointASNL: Robust Point Clouds Processing using Nonlocal Neural Networks with Adaptive Samplingcited by 239
  4. 2021PAConv: Position Adaptive Convolution with Dynamic Kernel Assembling on Point Cloudscited by 211
  5. 2021Walk in the Cloud: Learning Curves for Point Clouds Shape Analysiscited by 172
  6. 2018Generating 3D Adversarial Point Cloudscited by 144
  7. 2018PointCloud Saliency Mapscited by 112
  8. 2019PointCleanNet: Learning to Denoise and Remove Outliers from Dense Point Cloudscited by 99
  9. 2019Extending Adversarial Attacks and Defenses to Deep 3D Point Cloud Classifierscited by 92
  10. 2019AdvPC: Transferable Adversarial Perturbations on 3D Point Cloudscited by 84
  11. 2016Shape Completion Enabled Robotic Graspingcited by 79
  12. 2019SampleNet: Differentiable Point Cloud Samplingcited by 74
  13. 2020Learning Geometry-Disentangled Representation for Complementary Understanding of 3D Object Point Cloudcited by 71
  14. 2022Benchmarking and Analyzing Point Cloud Classification under Corruptionscited by 70
  15. 2021PointCutMix: Regularization Strategy for Point Cloud Classificationcited by 69
  16. 2019Adversarial Attack and Defense on Point Setscited by 63
  17. 2021Regularization Strategy for Point Cloud via Rigidly Mixed Samplecited by 59
  18. 2019Robustness of 3D Deep Learning in an Adversarial Settingcited by 55
  19. 2020LG-GAN: Label Guided Adversarial Network for Flexible Targeted Attack of Point Cloud-based Deep Networkscited by 55
  20. 2018A LiDAR Point Cloud Generator: from a Virtual World to Autonomous Drivingcited by 54
  21. 2021Less is More: Pay Less Attention in Vision Transformerscited by 53
  22. 2022Shape-invariant 3D Adversarial Point Cloudscited by 50
  23. 2021PointGuard: Provably Robust 3D Point Cloud Classificationcited by 49
  24. 2019PointHop: An Explainable Machine Learning Method for Point Cloud Classificationcited by 46
  25. 2020On Isometry Robustness of Deep 3D Point Cloud Models under Adversarial Attackscited by 46
  26. 2020IF-Defense: 3D Adversarial Point Cloud Defense via Implicit Function based Restorationcited by 46
  27. 2020Minimal Adversarial Examples for Deep Learning on 3D Point Cloudscited by 41
  28. 2021Point Cloud Augmentation with Weighted Local Transformationscited by 39
  29. 2017What does Attention in Neural Machine Translation Pay Attention to?cited by 30
  30. 2020ShapeAdv: Generating Shape-Aware Adversarial 3D Point Cloudscited by 20
  31. 2023Robustifying Token Attention for Vision Transformerscited by 19
  32. 2021Surrogate Model-Based Explainability Methods for Point Cloud NNscited by 18
  33. 20203D Point Cloud Feature Explanations Using Gradient-Based Methodscited by 14
  34. 2022LPF-Defense: 3D Adversarial Defense based on Frequency Analysiscited by 10
  35. 2021Explainability-Aware One Point Attack for Point Cloud Neural Networkscited by 7
  36. 2023PointCert: Point Cloud Classification with Deterministic Certified Robustness Guaranteescited by 6
  37. 2023EPiC: Ensemble of Partial Point Clouds for Robust Classificationcited by 5
  38. 2021CloudWalker: Random walks for 3D point cloud shape analysiscited by 4

50 most cited of 53 All 53 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2024: 1; 2025: 0; 2026: 1.

2 papers.

  1. 2024Enhancing Robustness to Noise Corruption for Point Cloud Recognition via Spatial Sorting and Set-Mixing Aggregation Modulecited by 0
  2. 2026Point-Selection Fine-Tuning Framework for Robust Point Cloud Classificationcited by 0