LEAR Research team
Learning and recognition in vision
Team presentationThe LEAR project-team is a research group hosted by INRIA Grenoble - Rhône-Alpes Research Centre in Montbonnot/Grenoble. It has been created in July 2003.
Our main research activities are object recognition and scene interpretation for static images and video sequences. These are amongst the most challenging problems in computer vision: it is today impossible to automatically determine the content of an image or a video sequence. We believe that the addition of learning techniques to computer vision will improve the current systems significantly. Even a partial solution to the problem will enable many applications. We are in particular interested in image retrieval and video indexing.
Research themesWe pursue three main areas of research:
- Image description.
Many efficient image description techniques are now available, such as for example affine interest points. We plan to extend these techniques in order to describe textures, to define significant similarity measures and to characterize 2D and 3D shape information.
Our main research area is computer vision. In the beginning, we will examine existing learning techniques and select those which show best performance in our context. We will then adapt learning theory and algorithms to take into account vision specific constraints.
- Building visual models.
To build visual models for recognizing objects, we propose to combine learning and image description techniques. For a given problem, the best learning and description technique has to be selected. In the beginning, this selection will done manually for each object class. We then plan to make this choice automatic. Partial automation is possible, for example based on feature selection.
Research teams of the same theme :
- AYIN - Stochastic models for remote sensing and skincare image processing
- MAGRIT - Visual Augmentation of Complex Environments
- MORPHEO - Capture and Analysis of Shapes in Motion
- PERCEPTION - Interpretation and Modelling of Images and Videos
- PRIMA - Perception, recognition and integration for observation of activity
- SIROCCO - Analysis representation, compression and communication of visual data
- STARS - Spatio-Temporal Activity Recognition Systems
- TEXMEX - Multimedia content-based indexing
- WILLOW - Models of visual object recognition and scene understanding