How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
Machine learning may frighten many social scientists, but developers are making major progress on creating computer software that is both self-managing and proactive. At the Neural Information Processing Systems Conference (NIPS2003) held in Vancouver, BC, Intel researchers announced that they were ready to release to the public a new version of open-source Probabilistic Networks Library (PNL). Offered through Intel?s Open Source Machine Learning Library (OpenML), the applications envisioned range from many aspects of computer vision to such complex computational tasks as data mining for biomedical to gene sequencing (see figure below). PNL is a library for inference and learning via graphical models.
After a limited distribution of alpha copies, PNL is now available at www.intel.com/research/mrl/pnl. While there are a large number of available software packages, the functionality of PNL is patterned after that of the Bayes Net Toolbox - but in C/C++ rather than in Matlab. What?s more, it was architected and written from scratch, according to Gary R. Bradski, manager of the Machine Learning Group at Intel Labs.
OpenML is analogous to OpenCV, Intel?s open source computer vision library, in that developers can look forward to a liberal licensing agreement that permits commercial use in certain circumstances. Furthermore, Intel benefits by gathering a tremendous amount of data for planning future architectures and products.
There have been numerous published papers about success with these open-source libraries, and many are concentrated in the field of computer vision. Vision vendors such as Visionics, Visage, Eyematic and Canesta are active participants in ML, as well as security firms like Intellivision. Of course, the most recognizable name on the ML list is Google - unless you are a Microsoft fan, that is.