Factorized Convolution with Spectral Normalization for Fundus Screening

Zeng，Ming1; Zeng，Na1; Fang，Jiansheng1,2; Liu，Jiang1

10.1109/ISBI52829.2022.9761526

2022

19th IEEE International Symposium on Biomedical Imaging (IEEE ISBI)

1945-7928

1945-8452

978-1-6654-2924-5

Proceedings - International Symposium on Biomedical Imaging

2022-March

1-5

28-31 March 2022

Kolkata, India

345 E 47TH ST, NEW YORK, NY 10017 USA

IEEE

Convolutional neural network (CNN) models have been widely used for fundus-based disease screening, but the model deployment is challenging due to the large demand for computing resources. The low-rank decomposition is usually used to compress CNN models. However, the reduction of model parameters often leads to performance degradation. Therefore, we propose a factorized convolution with spectral normalization, named FConvSN, to reduce the model complexity while maintaining an ideal performance. FConvSN applies the spectral norm to constrain the weight in the direction of the spectral norm to achieve weight decay, thereby improving the generalizability. Since the features of fundus images appear as a highly skewed distribution, factorized convolution can be used to promote the sharing of convolution parameters, and spectral normalization can further prevent excessive weight in the spectral norm direction. We have conducted experiments on the fundus dataset to prove that our FConvSN can achieve performance comparable to standard convolution.

Factorized convolution Fundus screening Spectral normalization

第一 ; 通讯

英语

CPCI-S ; EI

Engineering ; Radiology, Nuclear Medicine & Medical Imaging

Engineering, Biomedical ; Radiology, Nuclear Medicine & Medical Imaging

WOS:000836243800127

20221912089232

Convolutional neural networks ; Diagnosis

Medicine and Pharmacology:461.6 ; Information Theory and Signal Processing:716.1

2-s2.0-85129610955

Scopus

1.Southern University of Science and Technology,Department of Computer Science and Engineering,Shenzhen,China
2.Cvte Research,Guangzhou,China

Zeng，Ming,Zeng，Na,Fang，Jiansheng,et al. Factorized Convolution with Spectral Normalization for Fundus Screening[C]. 345 E 47TH ST, NEW YORK, NY 10017 USA:IEEE,2022:1-5.