CLASS, or Cognitive-Level Annotation using latent Statistical Structure, is a research initiative focussed on technologies to recognise three dimensional, physical objects such as a car, or lamp post, or a house, irrespective of its orientation, or the percentage of it visible to a machine vision system.

Status: Completed

Scientific Goals (From Project Site)

Class will develop a basic cognitive ability for use in intelligent content analysis: the automatic discovery of content categories and attributes from unstructured content streams. The demonstrators will focus on object recognition and scene analysis in images and video with accompanying text streams. Autonomous learning will make recognition more adaptive and allow more general classes and much larger and more varied data sets to be handled.

Technically, the work will combine latent structure models and semi-supervised learning methods from machine learning with advanced visual descriptors from computer vision and state-of-the-art text analysis techniques. Three levels of abstraction will be studied: new individuals (specific people, objects, scenes, actions); new object classes and attributes; and hierarchical categories and relations between entities.

Organisations Involved

*Note: For each organisation, a link to their website is provided, where known. If we have local information on the lab, company, or group then that information is also provided. This is standard layout on all Research Initiatives resources.

Details

The CLASS project, whilst academic in nature, succeeded in creating significant technological improvements to machine vision systems which have since been utilised in multiple commercial products for machine vision, from several different firms.

Chief amongst the advancements was a image splitting system that breaks any physical object up into a number of three dimensional patches. These patches are then compared against the object in question, both to see if the patch fits, and then to see if a selection of other patches also fit, without damaging the spatial relationship between the patches. Thus, partially occluded, or even damaged objects can still be recognised by the AI.

In addition, CLASS created special mechanisms - known as efficient approximate neighbourhood searches - for the comparison of an image or an object with huge numbers of reference images.

Project Home

CLASS Home Page

Dates

1st of January 2006 to 31st of December 2008

Funding

Total funding: 2.9 million Euros.

3 year 6th Framework Specific Targeted Research Project, European Union