Texture Bias

Texture bias, the tendency of deep learning models to prioritize texture over shape information when processing images, is a significant challenge hindering the generalization and robustness of computer vision systems. Current research focuses on mitigating this bias through various techniques, including data augmentation strategies (e.g., style transfer, edge enhancement), architectural modifications (e.g., incorporating transformers alongside convolutional neural networks), and novel training methods (e.g., curriculum learning, inductive bias distillation). Overcoming texture bias is crucial for improving the reliability and applicability of deep learning models across diverse domains and real-world scenarios, particularly in applications requiring robust object recognition and semantic understanding.

Papers

September 16, 2024

On Synthetic Texture Datasets: Challenges, Creation, and Curation
Blaine Hoak, Patrick McDaniel
Technical Challenge Synthetic Image Visual Creation Automatic Curation Texture Bias Texture Information

March 13, 2024

An Analysis of Human Alignment of Latent Diffusion Models
Lorenz Linhardt, Marco Morik, Sidney Bender, Naima Elosegui Borras
Diffusion Model General Analysis Latent Diffusion Model ImageNet 1k Representation Alignment Human Alignment Representation Bottleneck Texture Bias

February 14, 2024

Reducing Texture Bias of Deep Neural Networks via Edge Enhancing Diffusion
Edgar Heinert, Matthias Rottmann, Kira Maag, Karsten Kahl
Convolutional Neural Network Deep Neural Network Semantic Segmentation Texture Bias

January 17, 2024

Trapped in texture bias? A large scale comparison of deep instance segmentation
Johannes Theodoridis, Jessica Hofmann, Johannes Maucher, Andreas Schilling
Native Robustness Deep Learning Model Data Augmentation Pre Trained Model Instance Segmentation Texture Bias Large Scale Comparison

January 16, 2024

Deep Shape-Texture Statistics for Completely Blind Image Quality Evaluation
Yixuan Li, Peilin Chen, Hanwei Zhu, Keyan Ding, Leida Li, Shiqi Wang
Image Quality Blind Image Quality Assessment Texture Bias Image Statistic

October 11, 2023

Does resistance to style-transfer equal Global Shape Bias? Measuring network sensitivity to global shape configuration
Ziqi Wen, Tianqin Li, Zhi Jing, Tai Sing Lee
Style Transfer Network Sensitivity Global Shape Effective Resistance Texture Bias Shape Bias Global Sensitivity

August 18, 2023

Data augmentation and explainability for bias discovery and mitigation in deep learning
Agnieszka Mikołajczyk-Bareła
Deep Learning Data Augmentation High Explainability Faulty Negative Mitigation Bias Detection Texture Bias Bias Discovery

February 2, 2023

Domain Generalization Emerges from Dreaming
Hwan Heo, Youngjin Oh, Jaewon Lee, Hyunwoo J. Kim
Domain Generalization Generalization Performance Distribution Generalization Texture Bias

January 9, 2023

A Study on the Generality of Neural Network Structures for Monocular Depth Estimation
Jinwoo Bae, Kyumin Hwang, Sunghoon Im
Study Feature Monocular Depth Estimation CNN Transformer Texture Bias Neural Network Structure Shape Bias

November 23, 2022

BiasBed -- Rigorous Texture Bias Evaluation
Nikolai Kalischek, Rodrigo C. Daudt, Torben Peters, Reinhard Furrer, Jan D. Wegner, Konrad Schindler
Style Consistency Biased Training Biased Feature Texture Bias

November 14, 2022

Robustifying Deep Vision Models Through Shape Sensitization
Aditay Tripathi, Rishubh Singh, Anirban Chakraborty, Pradeep Shenoy
Deep Vision Model ImageNet E Texture Bias

November 2, 2022

Style Augmentation improves Medical Image Segmentation
Kevin Ginsburger
Medical Image Segmentation Segmentation Performance Texture Bias Style Augmentation Augmentation Approach

August 8, 2022

Exploiting Shape Cues for Weakly Supervised Semantic Segmentation
Sungpil Kho, Pilhyeon Lee, Wonyoung Lee, Minsong Ki, Hyeran Byun
Semantic Segmentation Segmentation Network Weakly Supervised Semantic Segmentation Texture Bias

June 12, 2022

InBiaseD: Inductive Bias Distillation to Improve Generalization and Robustness through Shape-awareness
Shruthi Gowda, Bahram Zonooz, Elahe Arani
Native Robustness Strong Generalization Inductive Bias Texture Bias

May 23, 2022

Deep Digging into the Generalization of Self-Supervised Monocular Depth Estimation
Jinwoo Bae, Sungho Moon, Sunghoon Im
Strong Generalization Monocular Depth Estimation Self Supervised Monocular Depth Estimation Texture Bias Shape Bias

April 25, 2022

Surpassing the Human Accuracy: Detecting Gallbladder Cancer from USG Images with Curriculum Learning
Soumen Basu, Mayank Gupta, Pratyaksha Rana, Pankaj Gupta, Chetan Arora
Image Classification State of the Art Curriculum Learning Texture Bias Ultrasound Image Analysis Human Accuracy Gallbladder Cancer Detection

February 16, 2022

A Developmentally-Inspired Examination of Shape versus Texture Bias in Machines
Alexa R. Tartaglini, Wai Keen Vong, Brenden M. Lake
Deep Neural Network New Machine Inductive Bias Object Shape Texture Bias Texture Feature Shape Bias

December 22, 2021

Recur, Attend or Convolve? On Whether Temporal Modeling Matters for Cross-Domain Robustness in Action Recognition
Sofia Broomé, Ernest Pokropek, Boyu Li, Hedvig Kjellström
Action Recognition Direct Convolution Temporal Modeling Temporal Difference Type Recurrence Texture Bias Cross Domain Robustness