Pith. sign in

REVIEW 19 cited by

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 1811.12231 v3 pith:Q7GN666P submitted 2018-11-29 cs.CV cs.AIcs.LGq-bio.NCstat.ML

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

classification cs.CV cs.AIcs.LGq-bio.NCstat.ML
keywords cnnshumanrepresentationtowardsbehaviouralbiasedimageimagenet
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Convolutional Neural Networks (CNNs) are commonly thought to recognise objects by learning increasingly complex representations of object shapes. Some recent studies suggest 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.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 19 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Geometric Collapse: When Vision Models Fail to Verify Physical Causality

    cs.CV 2026-07 accept novelty 7.0

    Modern depth predictors adopt physically unsupported edge cues as real geometry, causing global structural hallucinations that local repair cannot fix.

  2. Identifying Latent Concepts and Structures for Generalized Category Discovery

    cs.CV 2026-07 unverdicted novelty 7.0

    CPF-GCD enforces low-rank compositional structure on vision backbone features via spatial primitive fields so that novel categories emerge as new activation patterns over a shared vocabulary of reusable visual primitives.

  3. $A^2$: Smaller Self-Supervised ViTs Localize Better than Larger Ones

    cs.CV 2026-06 unverdicted novelty 7.0

    Smaller self-supervised ViTs localize objects better via attention than larger ViTs, enabling A² to decouple localization from feature extraction for competitive performance on distribution-shifted benchmarks.

  4. Cumulative Meta-Learning from Active Learning Queries for Robustness to Spurious Correlations

    cs.LG 2026-05 unverdicted novelty 7.0

    CAML meta-learns a progressively refined inductive bias from active-learning queries to improve robustness to spurious correlations, reporting accuracy gains on minority groups across several benchmarks.

  5. Dissociating spatial frequency reliance from adversarial robustness advantages in neurally guided deep convolutional neural networks

    q-bio.NC 2026-05 unverdicted novelty 7.0

    Spatial-frequency biases in neurally aligned DCNNs emerge from human-like representations but do not primarily drive their adversarial robustness advantages.

  6. Gaslight, Gatekeep, V1-V3: Early Visual Cortex Alignment Shields Vision-Language Models from Sycophantic Manipulation

    cs.CV 2026-04 conditional novelty 7.0

    Alignment of vision-language models with human V1-V3 early visual cortex negatively predicts resistance to sycophantic gaslighting attacks.

  7. Beyond Compression: Quantifying Spectral Accessibility in Vision Representations

    cs.CV 2026-06 unverdicted novelty 6.0

    Vision encoders alter spectral accessibility non-monotonically across depth with architecture-specific effects from projections and pooling, quantified via a new residual loss against random baselines.

  8. Almost for Free: Crafting Adversarial Examples with Convolutional Image Filters

    cs.LG 2026-05 conditional novelty 6.0

    Optimized 3x3 adversarial image filters based on edge detection generate transferable untargeted attacks on neural networks with 30-80% success using only one pass and far fewer parameters than prior methods.

  9. ShapeY: A Principled Framework for Measuring Shape Recognition Capacity via Nearest-Neighbor Matching

    cs.CV 2026-04 unverdicted novelty 6.0

    ShapeY is a benchmark dataset and nearest-neighbor protocol that measures shape-based recognition in vision models, revealing that even state-of-the-art networks fail to generalize consistently across 3D viewpoints an...

  10. Privacy Beyond Pixels: Latent Anonymization for Privacy-Preserving Video Understanding

    cs.CV 2025-11 conditional novelty 6.0

    A plug-and-play Anonymizing Adapter Module removes private information from video latent features using self-supervised privacy objectives and consistency losses while retaining utility on action recognition, temporal...

  11. A combination of noise and bilateral filters achieve supralinear and scalable adversarial robustness in CNNs

    cs.LG 2026-06 unverdicted novelty 5.0

    A preprocessor of Gaussian noise plus bilateral filtering yields supralinear adversarial robustness in CNNs and, when paired with adversarial training, ranks near the top of RobustBench while using far less compute, p...

  12. Dual-axis attribution of zebrafish tectal microcircuits for energy-efficient and robust neurocomputing

    cs.NE 2026-05 conditional novelty 5.0

    Zebrafish tectal subcircuits are dissociated into spike-efficient information gating and feedback-like robustness stabilization, then transferred to improve ResNet efficiency and noise tolerance.

  13. Optimization in Sparse 2D to Dense 3D Weakly Supervised Learning: Application to Multi-Label Segmentation of Large ex vivo MRI Data

    eess.IV 2026-05 unverdicted novelty 5.0

    Sparse-to-dense 3D segmentation from 2D slices shows divergent regularization needs: 2D benefits from strong augmentation and soft labels while 3D does not, and human-centric preprocessing harms performance.

  14. Accuracy Improvement of Cell Image Segmentation Using Feedback Former

    cs.CV 2024-08 unverdicted novelty 5.0

    Feedback Former improves cell image segmentation accuracy by feeding detailed feature maps back from near the output to lower transformer layers, outperforming non-feedback baselines with lower computational cost on t...

  15. Architectural Bias in Face Presentation Attack Detection: A Comparative Study of Vision Transformers and Convolutional Neural Networks

    cs.CV 2026-06 unverdicted novelty 4.0

    Pretrained DeiT-S Vision Transformer reaches 97.27% accuracy and cuts ethnic ACER gap to 0.13% on CeFA dataset while showing 3.6x better zero-shot generalization than ResNet18 CNN.

  16. Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture the full range of natural behavior

    q-bio.NC 2025-02 unverdicted novelty 4.0

    Advocates integrating naturalistic paradigms and AI progress into cognitive science to develop generalizable models of natural behavior while retaining experimental control and theoretical insight.

  17. Learning to Find Correlated Features by Maximizing Information Flow in Convolutional Neural Networks

    cs.CV 2019-06 unverdicted novelty 4.0

    Introduces IFM loss regularization for CNNs to learn correlated discriminative features, tested on shiftedMNIST dataset.

  18. Dynamic Computerized Tumbling-E Testing for Temporal Reliability of Human Sequential Perceptual Decisions

    q-bio.NC 2026-06 unverdicted novelty 3.0

    A dataset of 1,154 trials from 21 humans on a dynamic tumbling-E task reports mean reaction time of 1.5 s, 6.6% timeout rate, and smooth adaptation to small optotypes.

  19. Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture the full range of natural behavior

    q-bio.NC 2025-02 unverdicted novelty 3.0

    Position paper advocating integration of naturalistic paradigms and AI models to create generalizable theories of natural human behavior and cognition.