Computational Vision

Computer Vision Aids for the Visually Impaired - Text detection algorithm running on a mobile phone

We explore the multifaceted world of image and scene understanding combining computer vision and machine learning ingredients. Computer vision methods are used to extract information from the visual signals, while we resort to machine learning to model variability, and to gain robustness and flexibility. Within this framework we are pursuing research on object detection and object recognition as well as behavior analysis and action recognition. We devise biologically-inspired computational models of vision with a twofold aim: on one hand, to develop artificial vision systems that are able to face real-world action tasks, such as navigation and reaching, and on the other hand to exploit such models to better explain human perception.

Current projects