mlsanityData to 7 Oct 2026

A Developmentally-Inspired Examination of Shape versus Texture Bias in Machines

arXiv 2202.08340 ↗First posted 16 Feb 2022cs.CV cs.LGcited by 10

Not on the map. Why?

Abstract (arXiv)

Early in development, children learn to extend novel category labels to objects with the same shape, a phenomenon known as the shape bias. Inspired by these findings, Geirhos et al. (2019) examined whether deep neural networks show a shape or texture bias by constructing images with conflicting shape and texture cues. They found that convolutional neural networks strongly preferred to classify familiar objects based on texture as opposed to shape, suggesting a texture bias. However, there are a number of differences between how the networks were tested in this study versus how children are typically tested. In this work, we re-examine the inductive biases of neural networks by adapting the stimuli and procedure from Geirhos et al. (2019) to more closely follow the developmental paradigm and test on a wide range of pre-trained neural networks. Across three experiments, we find that deep neural networks exhibit a preference for shape rather than texture when tested under conditions that more closely replicate the developmental procedure.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 13 of its 31 references that are in the map. First posted on arXiv 16 Feb 2022.

Builds on

13 of 31 references.

References · 13

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  4. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  5. 2016Aggregated Residual Transformations for Deep Neural Networkscited by 2,540
  6. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  7. 2019Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNetcited by 260
  8. 2018Generalisation in humans and deep neural networkscited by 237
  9. 2021Partial success in closing the gap between human and machine visioncited by 185
  10. 2021Are Convolutional Neural Networks or Transformers more like human vision?cited by 109
  11. 2020Self-supervised learning through the eyes of a childcited by 61
  12. 2017Cognitive Psychology for Deep Neural Networks: A Shape Bias Case Studycited by 46
1 more
  1. 2018Learning Inductive Biases with Simple Neural Networkscited by 8

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2022: 2; 2023: 3; 2024: 1; 2025: 1; 2026: 3.

10 papers.

  1. 2023On the Foundations of Shortcut Learningcited by 64
  2. 2024When does compositional structure yield compositional generalization? A kernel theorycited by 19
  3. 2023MEWL: Few-shot multimodal word learning with referential uncertaintycited by 17
  4. 2022Representation Learning by Detecting Incorrect Location Embeddingscited by 5
  5. 2023Deep Neural Networks Can Learn Generalizable Same-Different Visual Relationscited by 4
  6. 2025Visual Anagrams Reveal Hidden Differences in Holistic Shape Processing Across Vision Modelscited by 3
  7. 2022A novel feature-scrambling approach reveals the capacity of convolutional neural networks to learn spatial relationscited by 1
  8. 2026On the Reliability of Cue Conflict and Beyondcited by 1
  9. 2026Balancing Frequencies and Pixels in Flow Matchingcited by 1
  10. 2026Feature, Alignment, and Supervision in Category Learning: A Comparative Approach with Children and Neural Networkscited by 0