Building an automatic sprite generator with deep convolutional generative adversarial networks

In recent years, image generation using Convolutional Neural Networks (CNNs) has become increasingly popular in the computer vision domain. However, there is less attention on using CNNs for sprite generation for games. A possible reason for this is that the amount of available sprite data in games is significantly less than in other domains, which typically use hundreds of thousands of images, or even more. In this work, we provide some beginning evidence that CNNs can be utilized for game-style sprite generation, even with small input datasets. We utilize a class of Generative Adversarial Network (GAN) known as a Deep Convolutional Generative Adversarial Network (DCGAN) for unsupervised learning and generation of new sprites. We have trained our network on various custom datasets, which contains human-like characters, faces and creatures in general. Results show evidence that CNNs can generate unique sprites from the input data that was provided as input.