Education, Science, Technology, Innovation and Life
Open Access
Sign In

Image Classification of Fashion-mnist Data Set Based on VGG Network

Download as PDF

DOI: 10.23977/iset.2019.004

Author(s)

Chao Duan, Panpan Yin, Yan Zhi, Xingxing Li

Corresponding Author

Yan Zhi

ABSTRACT

Image classification is the core of computer vision, which plays a pioneer role in other visual tasks. In this paper, we use VGG-11 network to classify Fashion-mnist data sets. We use several consecutive 3×3 convolution cores to replace the larger convolution cores (11×11, 7×7, 5×5) in AlexNet. For a given receptive field, small cumulative convolution kernel is better than large convolution kernel. Multilayer nonlinearity layer increases network depth to ensure learning more complex patterns, and the cost is relatively small. At the same time, batch normalization layer is added after each pooling layer, which makes it easier to train effective models by standardizing input data to make the distribution of each feature similar. Finally, the classification accuracy of this paper on Fashion-mnist Data Set is 91.5%.

KEYWORDS

Image Classification, VGG, Fashion-mnist, Batch Normalization

All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.