Abstract
Designing a logo for a new brand is a lengthy and tedious back-and-forthprocess between a designer and a client. In this paper we explore to whatextent machine learning can solve the creative task of the designer. For this,we build a dataset -- LLD -- of 600k+ logos crawled from the world wide web.Training Generative Adversarial Networks (GANs) for logo synthesis on suchmulti-modal data is not straightforward and results in mode collapse for somestate-of-the-art methods. We propose the use of synthetic labels obtainedthrough clustering to disentangle and stabilize GAN training. We are able togenerate a high diversity of plausible logos and we demonstrate latent spaceexploration techniques to ease the logo design task in an interactive manner.Moreover, we validate the proposed clustered GAN training on CIFAR 10,achieving state-of-the-art Inception scores when using synthetic labelsobtained via clustering the features of an ImageNet classifier. GANs can copewith multi-modal data by means of synthetic labels achieved through clustering,and our results show the creative potential of such techniques for logosynthesis and manipulation. Our dataset and models will be made publiclyavailable at https://data.vision.ee.ethz.ch/cvl/lld/.