The latest dlib 19.6 release has been announced for general availability from official sources. This release comes with several fixes, improvements, and new features as highlighted below.
The dlib library is a C++ toolkit which contains a collection of machine learning algorithms. It also contains tools for writing complex C++ based software especially those designed to solve real-world problems.
Highlights of the new dlib 19.6 Release
This version comes hot on the heels of dlib 19.5 release. It contains just a single major bug fix that was causing a build error in Visual Studio when CUDA is enabled.
Other features and improvements where released with version 19.5 as follows.
Support for cuDNN v6 and v7 has been included as is a Python wrapper for the CNN face detector. dlib models can now be converted to caffe models.
ARM NEON instructions are now supported making HOG based object detection faster on mobile devices.
There is now added support for calling C++ code from Java. There is a host of other new features in addition to these. Also is a long list of bug fixes and non-backward compatible changes that should not be overlooked. All this has been highlighted here on the official release notes page.
You can download a copy of the dlib 19.6 release library from the official release notes page here.