Home  |  Centre for Mathematical Sciences  |  LTH  |  LU

Autometa

In recent years the advancements in digital camera and storage technology has made photography easy and inexpensive. As a result even private image databases commonly includes thousands or tens of thousands of images. This creates a demand for efficient storing, categorizing and searching among images. The Autometa project aims at creating open solution for the exchange of information between mobile devices and between mobile devices and servers/home computers, and development of computer vision algorithms for generating such information. This will result in demo-applications that combines computer vision and organized distributed searches on mobile platforms. The idea is to generate information about the images, metadata, from the image content, and to store it in the images headers. By including information about the image in the image file itself, several possibilities for new functionality is made available. In a database context, searching for all images of a certain scene or including some object would be possible, in an online phase on a cellphone, taking a photograph of a certain person could automatically generate an RSS-feed to subscribers interested in pictures of that person. Performing much or all image analysis directly on the cell phone minimizes the need of sharing potentially sensitive information while still allowing a community supported recognition. In the case of face recognition, should a person not be recognized on the primary device, a "fingerprint" describing the face can be sent to other devices. Should they recognize the person, they may choose to share this information, or if the requesting party is unknown or untrusted keep it a secret.

Research is focused at Cognitive vision:
Detection and recognition of faces, scene categorization, object detection and localization.

Funded by VINOVA (The Swedish Governmental Agency for Innovation Systems).
Period: 2009-2011.
Partners:

  • Conveneer AB
  • Cognimatics AB
  • Sony Ericsson Mobile Communications AB