mlsanityData to 14 Oct 2026

Towards Understanding Adversarial Robustness of Optical Flow Networks

arXiv 2103.16255 ↗First posted 30 Mar 2021cs.CVcited by 15

Not on the map. Why?

Abstract (arXiv)

Recent work demonstrated the lack of robustness of optical flow networks to physical patch-based adversarial attacks. The possibility to physically attack a basic component of automotive systems is a reason for serious concerns. In this paper, we analyze the cause of the problem and show that the lack of robustness is rooted in the classical aperture problem of optical flow estimation in combination with bad choices in the details of the network architecture. We show how these mistakes can be rectified in order to make optical flow networks robust to physical patch-based attacks. Additionally, we take a look at global white-box attacks in the scope of optical flow. We find that targeted white-box attacks can be crafted to bias flow estimation models towards any desired output, but this requires access to the input images and model weights. However, in the case of universal attacks, we find that optical flow networks are robust. Code is available at https://github.com/lmb-freiburg/understanding_flow_robustness.

Key facts. Cited by 15 papers in the map (data through 14 Oct 2026). Builds on 24 of its 39 references that are in the map. First posted on arXiv 30 Mar 2021.

Builds on

24 of 39 references.

References · 24

  1. 2014Explaining and Harnessing Adversarial Examplescited by 7,364
  2. 2013Intriguing properties of neural networkscited by 5,155
  3. 2015Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classificationcited by 3,611
  4. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
  5. 2020RAFT: Recurrent All-Pairs Field Transforms for Optical Flowcited by 2,415
  6. 2019Natural Adversarial Examplescited by 1,408
  7. 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,268
  8. 2015FlowNet: Learning Optical Flow with Convolutional Networkscited by 1,059
  9. 2017PWC-Net: CNNs for Optical Flow Using Pyramid, Warping, and Cost Volumecited by 968
  10. 2016Adversarial Machine Learning at Scalecited by 936
  11. 2015A Large Dataset to Train Convolutional Networks for Disparity, Optical Flow, and Scene Flow Estimationcited by 881
  12. 2016Universal adversarial perturbationscited by 784
12 more
  1. 2016FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networkscited by 784
  2. 2017One pixel attack for fooling deep neural networkscited by 582
  3. 2017Adversarial Patchcited by 517
  4. 2016Optical Flow Estimation using a Spatial Pyramid Networkcited by 424
  5. 2019Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Comingcited by 336
  6. 2019Adversarial Attacks and Defenses in Images, Graphs and Text: A Reviewcited by 237
  7. 2017Super-Convergence: Very Fast Training of Neural Networks Using Large Learning Ratescited by 199
  8. 2018Models Matter, So Does Training: An Empirical Study of CNNs for Optical Flow Estimationcited by 94
  9. 2017Universal Adversarial Perturbations Against Semantic Image Segmentationcited by 92
  10. 2019Attacking Optical Flowcited by 41
  11. 2018On the Robustness of the CVPR 2018 White-Box Adversarial Example Defensescited by 40
  12. 2020Stereopagnosia: Fooling Stereo Networks with Adversarial Perturbationscited by 19

Built on by

Cited by year

Citing papers per year: 2021: 1; 2022: 2; 2023: 5; 2024: 3; 2025: 4.

15 papers.

Sort:

By area:

  1. 2025Towards Understanding Subliminal Learning: When and How Hidden Biases Transfercited by 34
  2. 2023CosPGD: an efficient white-box adversarial attack for pixel-wise prediction taskscited by 27
  3. 2022A Perturbation-Constrained Adversarial Attack for Evaluating the Robustness of Optical Flowcited by 16
  4. 2023Distracting Downpour: Adversarial Weather Attacks for Motion Estimationcited by 15
  5. 2023DistractFlow: Improving Optical Flow Estimation via Realistic Distractions and Pseudo-Labelingcited by 14
  6. 2022Attacking Motion Estimation with Adversarial Snowcited by 9
  7. 2025When and How Does CLIP Enable Domain and Compositional Generalization?cited by 9
  8. 2023DIFT: Dynamic Iterative Field Transforms for Memory Efficient Optical Flowcited by 7
  9. 2025RobustSpring: Benchmarking Robustness to Image Corruptions for Optical Flow, Scene Flow and Stereocited by 7
  10. 2023Detection Defenses: An Empty Promise against Adversarial Patch Attacks on Optical Flowcited by 6
  11. 2024Beware of Aliases -- Signal Preservation is Crucial for Robust Image Restorationcited by 6
  12. 2021Consistent Semantic Attacks on Optical Flowcited by 4
3 more
  1. 2024Towards Dependable Optical Flow: Benchmarks for Real-World Spatiotemporal Corruptionscited by 3
  2. 2024Left-right Discrepancy for Adversarial Attack on Stereo Networkscited by 2
  3. 2025Test-Time Defense Against Adversarial Attacks via Stochastic Resonance of Latent Ensemblescited by 2