Textile Design Generation Using GANs

In this work, we propose a novel method for automated textile design patterns generation using generative models. We first improve the accuracy of state-of-the-art results in classification of textile design patterns by 2% through data cleaning and pseudo labeling. Then a new dataset which is an improvement of existing dataset is also proposed. On this new dataset we compare the performance of image generative models like Wasserstein Generative Adversarial Networks Gradient Penalty (WGANs GP), Deep Convolutional GANs (DCGANs) and Convolutional Variational Autoencoders (CVAEs) for all classes separately and have evaluated the models using the inception score. We further use a style transfer model to combine multiple designs generated by WGANs GP, due to its better results among all three approaches and form more complex and appealing textile designs. Moreover we present results of unsupervised clustering of different patterns in the latent space captured by a CVAE.