Generating Logos with Neural Networks / Spring 2019 - Master Thesis
Generative Adverserial Networks were succesfully applied to many image genration tasks with many extensions and variations. Among them is the Text to Image synthesis network . We would like to train a model for generating company logos given short keyword descriptors.
The project would involve constructing a training dataset consisting of pairs of keyword sequences and logo images. The data can be obtained either from DBPedia or by scraping Wikipedia. After this, the main task would be to research potential models, select promising one(s) and procede to implementation, training and evaluation.
Supervised by Bojan Karlas
-  Reed, S., Akata, Z., Yan, X., Logeswaran, L., Schiele, B. and Lee, H., 2016. Generative adversarial text to image synthesis. arXiv preprint arXiv:1605.05396. [arxiv]