Research on Extended Image Data Set Based on Deep Convolution Generative Adversarial Network
Zixi Liu, Ming Tong, Xiaoyu Liu, Zhixiong Du, Weicong Chen · 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC) · 2020
The Deep Convolution Generative Adversarial Network (DCGAN) adds the structure of Generative Adversarial Network (GAN) on the basis of the generation countermeasure network, and specially generates image samples. In this paper, DCGAN is used to generate the image which does not belong to MNIST data set, and then, a new data set is obtained. Finally, Convolutional Neural Networks (CNN) [1] is used to test the new data set. We need define an initializer to make the GAN converge better, and use the standard LEAKYRELU function to activate the GAN. The generator is defined by a fully connected layer with an input size of 128. The noise is Gaussian white noise, which uses RELU as the activation function. The discriminator uses two convolution layers, the first one uses RELU as the activation function, and the second layer uses sigmoid function. The results show that the accuracy of the new data set is the same as that of the original data set when tested on CNN, and the method of expanding MNIST data set by using deep convolution is effective.