APSIPA Transactions on Signal and Information Processing > Vol 12 > Issue 1

CNN Pretrained Model with Shape Bias using Image Decomposition

Akinori Iwata, Doshisha University, Japan, iwata@mm.doshisha.ac.jp , Masahiro Okuda, Doshisha University, Japan
 
Suggested Citation
Akinori Iwata and Masahiro Okuda (2023), "CNN Pretrained Model with Shape Bias using Image Decomposition", APSIPA Transactions on Signal and Information Processing: Vol. 12: No. 1, e42. http://dx.doi.org/10.1561/116.00000113

Publication Date: 25 Oct 2023
© 2023 A. Iwata and M. Okuda
 
Subjects
 
Keywords
CNNshape biasimage recognitionshape-dominant images
 

Share

Open Access

This is published under the terms of CC BY-NC.

Downloaded: 738 times

In this article:
Introduction 
Related Work 
Proposed Method 
Experiments 
Conclusion 
References 

Abstract

It is known that various implicit bias occur in Neural Networks due to their structural restrictions. Among them, texture bias caused by the convolution of CNNs has a significant impact on recognition performance. This paper shows that models with strong texture bias degrade recognition performance on datasets with large shape features, and to compensate for this characteristic of CNNs we introduce a method to increase their bias toward shapes rather than textures. Our method uses a simple image decomposition technique to create a shape-dominant dataset and then build a model with shape bias using the dataset. We experimentally show that the network can be biased towards shape without a significant loss of recognition accuracy compared to CNNs trained using conventional ImageNet. Additionally, we demonstrate that the CNN built by the proposed method obtains a higher recognition accuracy for shape-dominant images than those created using conventional methods.

DOI:10.1561/116.00000113