Cleant is a Polish clothing company, created by three brothers, that above all dotes on its community. Being close to its customers has been in the brand’s DNA since the very beginning. Trying to listen to all fans, Cleant decided to take its brand one step further and redefine its identity in a way all fans know they’re heard out.
The idea was that the whole community would create the new brand symbol, thereby becoming an inherent part of it. It has been achieved with the help of the newest technologies, more precisely Artificial Intelligence, powered naturally by NeuroSYS. Anyone could download a Cleant mobile app, design their project, and upload it. Then, an AI system developed specifically for this purpose would analyze all logos sent, find patterns, and generate a final sign encompassing the collective vision of how a Cleant logo should look.
This was the first visual identity created live by the brand community using AI. The unpredictability of the process excited tremendous attention. At some point, Cleant app was in 4th place in the AppStore in the entertainment category, just behind Tik Tok or Netflix.
Cleant redefines its visual identity in an innovative way, taking advantage of modern technology. The new corporate logo has been generated by Artificial Intelligence, more precisely Generative Adversarial Network (GAN). We had tested different types of GANs to verify their possibilities before we’ve chosen the final solution.
The rebranding project was run by Redkroft studio and we were responsible for providing the technological part – collecting fans’ images and generating the final logo with AI mechanisms. To do so, we’ve developed an AI system, namely a Generative Adversarial Network (GAN) – a deep neural network. We had verified the possibility of different types of GANs usage to generate a new logo based on a small amount of data (small compared to other projects, such as object detection in microbiology). The algorithms had to learn to find patterns in our dataset to finally give rise to a logo resulting from thousands of symbols drawn by Cleant fans.
It wouldn’t be possible without the technologies developed by our Research & Development team. We used GANs, an unsupervised learning technique, which consists of two networks: a generator, which learns to generate new logo samples, and a discriminator which is taught to recognize real examples from generated ones to improve the logo generation process. Additionally, to stabilize network training for limited amounts of data we used an adaptive discriminator augmentation (ADA) mechanism. As far as technologies are concerned, we’ve picked Python and a PyTorch library.
