University of Cambridge > > Microsoft Research Cambridge, public talks > Part Detection and Species Identification

Part Detection and Species Identification

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Microsoft Research Cambridge Talks Admins.

This event may be recorded and made available internally or externally via Microsoft will own the copyright of any recordings made. If you do not wish to have your image/voice recorded please consider this before attending

Species identification is an interesting example of fine-grained classification, in which an object must be assigned to one of a large number of very similar classes. It is also a problem with many practical applications that evokes great interest in the general public. I will describe a series of work that we have done aimed at using automatic recognition in field guides for species identification in a variety of domains.

One of the core technical problems in fine-grained classification is the identification of the parts of animals. This allows us to make meaningful comparisons between objects from similar animals. I will first describe our work on part detection using a non-parametric model of part configurations called a consensus of exemplars. We have applied this to the detection of fiducial points on human faces, and to the detection of animal parts. Then I will explain how these parts can be used for fine-grained classification of animal species. We experiment with this approach using dog breeds as a model problem. The resulting system is displayed in our iphone app Dogsnap, which uses visual classification to determine the breed of dog in a photograph. Similar ideas have also been used to develop Birdsnap, which also uses species distribution data to provide a practical field guide to birds. Finally, I will describe our earlier work developing Leafsnap, the first mobile app for identifying plant species using automatic visual recognition. Leafsnap has been downloaded by over a million users and has been used in many classrooms and in biodiversity studies.

This work has been done in collaboration with many people at Columbia University and the Smithsonian Institution National Museum of Natural History

This talk is part of the Microsoft Research Cambridge, public talks series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity