Generative Adversarial Networks was introduced by Ian Goodfellow in 2014 [1]. Huge improvements have been made in the technology since then.
This is my attempt, to create logos using GANs. I have tried DC GANs (Deep Convolutional Generative Adversarial Networks). However, there has not been any significant progress! A similar approach of DC GAN produced good results on the MNIST datasets. MAy be it is due to the higher complexity of the data and lesser number of samples that DC GANs does not produce the desirable results. I am hopeful and now trying Style-GAN [2] for the logo generation. Style GAN was developed at NVidia.
The data used is the intellectual property of their respective owners. I have no written/verbal consent from the owners of the images used. I obtained these images from Google Images. I do not wish to monetize the project in any form. The purpose of the project is purely academic and learning based. I hope I can use the data in a good light. If you want to use the data, use it at your own risk! There are 2019 images of various logos in the dataset. Each image is of the dimension 112 x 112 x 3.
- Goodfellow, Ian, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua Bengio. "Generative adversarial nets." In Advances in neural information processing systems, pp. 2672-2680. 2014.
- Karras, Tero, Samuli Laine, and Timo Aila. "A style-based generator architecture for generative adversarial networks." In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 4401-4410. 2019.