03 July 2012, 09:48

Computer vision library ccv reaches 0.1 milestone

Pedestrian detection with ccv
Source: libccv.org After two years in development, ccv 0.1, "a modern computer vision library", has been released. Ccv began development in 2010 when author Liu Liu, frustrated by problems with image preprocessing for a gesture recognition demonstration, decided to work on a different approach.

That approach was to implement a "handful" of state of the art algorithms and allow applications to construct their own processing paths. This allows developers using the library to skip the preprocessing layers often embedded in other libraries. The developers have said that libraries such as OpenCV offer great quality algorithms but with their current path of development are becoming frameworks; ccv's approach is, instead, to "strive to be a drop-in statically-linked library".

The algorithms include a fast detection of rigid objects, detection of "difficult" objects, text detection and invariant keypoint detection. The algorithms are brought together using a shared cache component and the design of the library makes them easy to embed in other applications. The project is available to download as a zip file or tar file, and can be forked from its GitHub repository.