High Resolution Generative Design

Nowadays generative methods that produce novel samples from high-dimensional data distributions, such as images, are finding widespread use thanks to recent developments in the field of Deep Learning. In particular, the use of increasingly complex generative models such as GANs (Generative Adversarial Networks) or Variational Autoencoders has allowed us to obtain exceptional results that were impossible only a few years ago. 

The world of design (especially in the industrial field) has begun to observe the development of these new technologies with great interest, understanding their high potential for various applications in their specific sector. It is from this interest that the collaboration between the AImageLab laboratory and the DigitalDesign Srl. design studio was born. The idea is to develop a customized generative model (in particular a GAN) capable of producing high resolution patterns for design surfaces, inspired by natural materials or abstract textures. 

The results obtained in the first phase of the project were certainly encouraging and prompted DigitalDesign to patent the solution (patent accepted) and to continue the collaboration with AImageLab for a second phase of the project, aimed at a further improvement of the generative system and the implementation of new useful features, such as image quality enhancement and color space conversion. Phase 2 of the project, still in progress, has already led to the filing of a second patent which is currently under review (patent pending).