Stable Diffusion: a model for generating images from text

In 2009, Fei-Fei Li presented ImageNet as a database with 1000 classes, each with 1000 images, at the CVPR. The database was created to provide the Computer Vision community with a data standard. In 2010, the ILSVRC competition was launched using ImageNet to evaluate models and algorithms. In 2012, a neural network, AlexNet, won this competition for the first time, marking a milestone in AI. Since then, there has been significant progress with examples such as ChatGPT and generative image models as its leading exponents.

In this talk we will discuss the technologies and concepts that have yielded the Stable Diffusion model, which can generate images from text, and we will explore both its capabilities and limitations.

Programa

Historical framework of generative image models
Technologies and concepts behind the Stable Diffusion model
Advantages and disadvantages of the model
Applications

Masterclass taught by the CVC (CIDAI core partner)

Play video Download Presentation

Taught by:

Coen Antes

Head of the Innovation Unit at the CVC

Coen studied Computer Science at Eindhoven University of Technology in the Netherlands. To round off his education, he took part in the “Mathematics for Industry” programme at the same University and to finish his master’s degree he spent six months at the Fraunhofer Institute for Mathematics for Industry in Kaiserslautern. After university, he worked at a Machine Vision company called Beltech where for three years he delivered computer vision solutions for industrial settings.
He then decided he needed a change and so he became a Research Support Engineer at the Computer Vision Center in Barcelona. After several years as an engineer and programmer, he is now head of the Innovation Unit at the CVC.