The FIRE-ID project considers the semantic annotation of visual content, such as photos or videos shared on social networks, or images captured by video surveillance devices or scanned documents. More specifically, the project considers the fine-grained recognition problem, where the number of classes is large and where classes are visually similar, for instance animals, products, vehicles or document forms. We also assumed that the amount of annotated data available per class for the learning stage is limited.
The project has started in May 2012 and will end in April 2015.
We are supporting the Micro-Workshop on Computer Vision at Inria Rennes on October 2nd 2014. The invited speakers are: Yannis Avrithis (NTUA), Albert Gordo (Xerox), Jiri Matas (CTUP), Patrick Pérez (Technicolor), Andrew Zisserman (University of Oxford).
Large-Scale Visual Recognition at CVPR‘2014. Speakers: Zaid Harchaoui (INRIA Grenoble), Hervé Jégou (INRIA Rennes) and Florent Perronnin (Xerox).
Visual Recognition in Large Collections at BMVC‘14, April 2014. Speaker: Hervé Jégou (INRIA Rennes).
Large-Scale Visual Recognition at CVPR‘13, June 2013. Speakers: Ondrej Chum (CVUT), Zaid Harchaoui (INRIA Grenoble), Hervé Jégou (INRIA Rennes), Florent Perronnin (Xerox), Marc’Aurelio Ranzato (Google) and Andrea Vedaldi (Oxford University).
Large-Scale Image Retrieval and Classification at CVPR‘12, June 2012. Speakers: Hervé Jégou (INRIA Rennes), Florent Perronnin (Xerox).
Our solution won the FGComp evaluation campaign for fine-grained image classification systems. The team consisted of: Philippe-Henri Gosselin (INRIA Rennes, ENSEA), Naila Murray (Xerox), Hervé Jégou (INRIA Rennes) and Florent Perronnin (Xerox).