Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System
The convolutional neural network (CNN) was initially inspired by the physiological visual system, and its structure has become increasingly complex after decades of development. Although CNN architectures now have diverged from biological structures, we believe that the mechanism of feature extracti...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-01-01
|
| Series: | Mathematics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2227-7390/13/1/142 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841549141322235904 |
|---|---|
| author | Bin Li Yuki Todo Sichen Tao Cheng Tang Yu Wang |
| author_facet | Bin Li Yuki Todo Sichen Tao Cheng Tang Yu Wang |
| author_sort | Bin Li |
| collection | DOAJ |
| description | The convolutional neural network (CNN) was initially inspired by the physiological visual system, and its structure has become increasingly complex after decades of development. Although CNN architectures now have diverged from biological structures, we believe that the mechanism of feature extraction in the visual system can still provide valuable insights for enhancing CNN robustness and stability. In this study, we investigate the mechanism of neuron orientation selectivity and develop an artificial visual system (AVS) referring to the structure of the primary visual system. Through learning on an artificial object orientation dataset, AVS acquires orientation extraction capabilities. Subsequently, we employ the pre-trained AVS as an information pre-processing block at the front of CNNs to regulate their preference for different image features during training. We conducted a comprehensive evaluation of the AVS–CNN framework across different image tasks. Extensive results demonstrated that the CNNs enhanced by AVS exhibit significant model stability enhancement and error rate decrease on noise data. We propose that incorporating biological structures into CNN design still holds great potential for improving overall performance. |
| format | Article |
| id | doaj-art-5d8e77995b0a456189af5be315ff3d0b |
| institution | Kabale University |
| issn | 2227-7390 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Mathematics |
| spelling | doaj-art-5d8e77995b0a456189af5be315ff3d0b2025-01-10T13:18:23ZengMDPI AGMathematics2227-73902025-01-0113114210.3390/math13010142Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual SystemBin Li0Yuki Todo1Sichen Tao2Cheng Tang3Yu Wang4Division of Electrical Engineering and Computer Science, Kanazawa University, Kanazawa 920-1192, JapanFaculty of Electrical, Information and Communication Engineering, Kanazawa University, Kanazawa 920-1192, JapanFaculty of Engineering, Toyama University, Gofuku, Toyama 930-8555, JapanFaculty of Information Science and Electrical Engineering, Kyushu University, Fukuoka 819-0395, JapanDivision of Electrical Engineering and Computer Science, Kanazawa University, Kanazawa 920-1192, JapanThe convolutional neural network (CNN) was initially inspired by the physiological visual system, and its structure has become increasingly complex after decades of development. Although CNN architectures now have diverged from biological structures, we believe that the mechanism of feature extraction in the visual system can still provide valuable insights for enhancing CNN robustness and stability. In this study, we investigate the mechanism of neuron orientation selectivity and develop an artificial visual system (AVS) referring to the structure of the primary visual system. Through learning on an artificial object orientation dataset, AVS acquires orientation extraction capabilities. Subsequently, we employ the pre-trained AVS as an information pre-processing block at the front of CNNs to regulate their preference for different image features during training. We conducted a comprehensive evaluation of the AVS–CNN framework across different image tasks. Extensive results demonstrated that the CNNs enhanced by AVS exhibit significant model stability enhancement and error rate decrease on noise data. We propose that incorporating biological structures into CNN design still holds great potential for improving overall performance.https://www.mdpi.com/2227-7390/13/1/142CNNrobustnessartificial visual system |
| spellingShingle | Bin Li Yuki Todo Sichen Tao Cheng Tang Yu Wang Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System Mathematics CNN robustness artificial visual system |
| title | Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System |
| title_full | Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System |
| title_fullStr | Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System |
| title_full_unstemmed | Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System |
| title_short | Enhancing Convolutional Neural Network Robustness Against Image Noise via an Artificial Visual System |
| title_sort | enhancing convolutional neural network robustness against image noise via an artificial visual system |
| topic | CNN robustness artificial visual system |
| url | https://www.mdpi.com/2227-7390/13/1/142 |
| work_keys_str_mv | AT binli enhancingconvolutionalneuralnetworkrobustnessagainstimagenoiseviaanartificialvisualsystem AT yukitodo enhancingconvolutionalneuralnetworkrobustnessagainstimagenoiseviaanartificialvisualsystem AT sichentao enhancingconvolutionalneuralnetworkrobustnessagainstimagenoiseviaanartificialvisualsystem AT chengtang enhancingconvolutionalneuralnetworkrobustnessagainstimagenoiseviaanartificialvisualsystem AT yuwang enhancingconvolutionalneuralnetworkrobustnessagainstimagenoiseviaanartificialvisualsystem |