mloss

At http://mloss.org/software/ it is possible to find a good collection of Open Source Machine Learning software. In my experience, this kind of resource is very valuable since the availability of good software is fundamental both not to "reinvent the wheel", and to have a quick "bootstrap" into the real problems of machine learning that are not connected with implementation of algorithms that are, by themselves, quite complex both in theory and for numerical considerations.

I gave a look quickly since I'm not needing any new algorithm now, and I saw some good projects that I had the opportunity to use in my past work:

• OpenCV for image processing;
• LibDAI for graph inference;

in addition to other things that I didn't use but I know by fame.