The field of artificial intelligence emerged at the intersection of mathematics, physics, computer science and cognitive science, with the quest of creating computer programs that learn from experience. The challenge is to build robots able to perceive, see the real world, and act intelligently in unexpected, uncertain circumstances, as humans do. Computational visual perception and motion capture technologies started more than a century ago with Helmholtz's mathematics of the eye and the emerging photographic techniques. Nowadays recognizing visual objects, understanding video content and transferring this into 3-dimensional models is the basis of advanced special effects, digital libraries and image indexing systems, or the construction of humanoid robots that can localize objects, recognize people, comprehend actions and interact with the world seamlessly. Our research group develops new computational methods for artificial intelligence and visual recognition, aiming at creating programs able to understand the visual world from images and at building mathematical models that can be fitted into the `mind' of an intelligent agent. Recent advances in visual recognition and learning from massive data collections makes this quest more exciting than ever before.