DeepAI AI Chat
Log In Sign Up

ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness

by   Robert Geirhos, et al.
Universität Tübingen

Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies hint to a more important role of image textures. We here put these conflicting hypotheses to a quantitative test by evaluating CNNs and human observers on images with a texture-shape cue conflict. We show that ImageNet-trained CNNs are strongly biased towards recognising textures rather than shapes, which is in stark contrast to human behavioural evidence and reveals fundamentally different classification strategies. We then demonstrate that the same standard architecture (ResNet-50) that learns a texture-based representation on ImageNet is able to learn a shape-based representation instead when trained on "Stylized-ImageNet", a stylized version of ImageNet. This provides a much better fit for human behavioural performance in our well-controlled psychophysical lab setting (nine experiments totalling 48,560 psychophysical trials across 97 observers) and comes with a number of unexpected emergent benefits such as improved object detection performance and previously unseen robustness towards a wide range of image distortions, highlighting advantages of a shape-based representation.


page 1

page 3

page 4

page 16

page 17

page 18


Does enhanced shape bias improve neural network robustness to common corruptions?

Convolutional neural networks (CNNs) learn to extract representations of...

Universal Adversarial Perturbations to Understand Robustness of Texture vs. Shape-biased Training

Convolutional Neural Networks (CNNs) used on image classification tasks ...

Increasing Shape Bias in ImageNet-Trained Networks Using Transfer Learning and Domain-Adversarial Methods

Convolutional Neural Networks (CNNs) have become the state-of-the-art me...

Interpreting Adversarially Trained Convolutional Neural Networks

We attempt to interpret how adversarially trained convolutional neural n...

Shape or Texture: Understanding Discriminative Features in CNNs

Contrasting the previous evidence that neurons in the later layers of a ...

CognitiveCNN: Mimicking Human Cognitive Models to resolve Texture-Shape Bias

Recent works demonstrate the texture bias in Convolutional Neural Networ...

Code Repositories


Code and models for the paper Shape-Texture Debiased Neural Network Training (ICLR 2021)

view repo


Code to create Stylized-ImageNet, a stylized version of standard ImageNet.

view repo


This code generates image datasets with annotations and labels for training your neural network by one click.

view repo


Code for my Master's Thesis: "The Role of Local Versus Global Features in Convolutional Neural Networks"

view repo