Longer message follows:
After nearly 1 year of development time, I'm very happy to announce to you the next generation of our Bob library for signal processing, machine learning and, of course, biometrics!
Version 2.0 contains an overhauled structure based on your feedback and our experience using and deploying Bob for our own research activities.
The first thing you'll notice is we are expanding installation possibilities by hosting the entirety of Bob on PyPI. This was possible thanks to a re-worked set of native Python bindings, which shall, in the near future, allows us to expand even further the number of supported features, external packages, operating systems and architectures.
So that you can focus on less dependencies and more on your problem, we have split Bob into smaller toolboxes with specialised functions. By combining one or more of those packages, you can easily create your own work environment, suiting your specific needs.
Don't worry! As always, we continue to strive in quality with our nightly builds where all packages are checked against each other:
Here is the current list of all the packages using the Bob namespace:
As you'll notice, we now keep a even tighter control of the build and its quality not only globally, but also on a per-package basis. This should translate in less bugs and better quality control over time.
Want to give Bob v2 a spin? Here you'll find our new installation instructions:
If you have the time, please drop us an e-mail in this mailing list sharing your experience.
Reporting bugs should go through the individual packages. If unsure, please use our centralised issue tracker:
Finally, I'd like to thank the whole of the Idiap Biometrics team for working hard on this new version of Bob - it was far from easy to get all going. A special thanks goes to Manuel Günther who did not rest while there was work to do for this release.
Enjoy, André