mlsanityData to 14 Oct 2026

Extreme Image Transformations Affect Humans and Machines Differently

arXiv 2212.13967 ↗First posted 30 Nov 2022cs.CV cs.AI cs.LGcited by 4

Not on the map. Why?

Abstract (arXiv)

Some recent artificial neural networks (ANNs) claim to model aspects of primate neural and human performance data. Their success in object recognition is, however, dependent on exploiting low-level features for solving visual tasks in a way that humans do not. As a result, out-of-distribution or adversarial input is often challenging for ANNs. Humans instead learn abstract patterns and are mostly unaffected by many extreme image distortions. We introduce a set of novel image transforms inspired by neurophysiological findings and evaluate humans and ANNs on an object recognition task. We show that machines perform better than humans for certain transforms and struggle to perform at par with humans on others that are easy for humans. We quantify the differences in accuracy for humans and machines and find a ranking of difficulty for our transforms for human data. We also suggest how certain characteristics of human visual processing can be adapted to improve the performance of ANNs for our difficult-for-machines transforms.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 24 of its 96 references that are in the map. First posted on arXiv 30 Nov 2022.

Builds on

24 of 96 references.

References · 24

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  3. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,448
  4. 2015Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classificationcited by 3,611
  5. 2020Shortcut Learning in Deep Neural Networkscited by 1,969
  6. 2021MLP-Mixer: An all-MLP Architecture for Visioncited by 1,446
  7. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,432
  8. 2019Self-training with Noisy Student improves ImageNet classificationcited by 1,311
  9. 2014Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Imagescited by 776
  10. 2019Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNetcited by 260
  11. 2018Generalisation in humans and deep neural networkscited by 237
  12. 2017Towards Robust Neural Networks via Random Self-ensemblecited by 159
12 more
  1. 2021Learning to See by Looking at Noisecited by 92
  2. 2021Robust and Accurate Object Detection via Adversarial Learningcited by 42
  3. 2019Putting visual object recognition in contextcited by 33
  4. 2018Humans can decipher adversarial imagescited by 31
  5. 2022Towards Robust Adaptive Object Detection under Noisy Annotationscited by 20
  6. 2019Noise Robust Generative Adversarial Networkscited by 17
  7. 2020IntegralAction: Pose-driven Feature Integration for Robust Human Action Recognition in Videoscited by 12
  8. 2021Tracking Without Re-recognition in Humans and Machinescited by 12
  9. 2021Neural Population Geometry Reveals the Role of Stochasticity in Robust Perceptioncited by 12
  10. 2020Deep-CAPTCHA: a deep learning based CAPTCHA solver for vulnerability assessmentcited by 11
  11. 2021The Challenge of Appearance-Free Object Tracking with Feedforward Neural Networkscited by 3
  12. 2022Robustness of Humans and Machines on Object Recognition with Extreme Image Transformationscited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 1; 2024: 0; 2025: 2; 2026: 1.

4 papers.

  1. 2025Towards Experience-Centered AI: A Framework for Integrating Lived Experience in Design and Developmentcited by 1
  2. 2023Extreme Image Transformations Facilitate Robust Latent Object Representationscited by 0
  3. 2025Stretching Beyond the Obvious: A Gradient-Free Framework to Unveil the Hidden Landscape of Visual Invariancecited by 0
  4. 2026Do Machines Fail Like Humans? A Human-Centred Out-of-Distribution Spectrum for Mapping Error Alignmentcited by 0