Proposing the Development of Dataset of Cartoon Character using DCGAN Model

In this paper, we propose to create animated images using the deep convolutional generative adversarial network (DCGAN) model for creative design of publications for entertainment applications. We have built a set of image data for machine learning with nearly 1000 images in order to increase their quality while reducing the machine learning execution time. The results show that the proposal model not only improves the accuracy of the image with 21 dB but also reduces the execution time to less than a week (with transposed convolution up to \(stride = 2, kernel = 3\) and size of image from \(64\, \times \,64 \,\times \,32\) to \(128 \,\times \,128 \,\times \,3\)). The results meet for requirements of future real-time applications.