A collection of tools to analyze tongue surface contours in ultrasound images.
##Matlab-version The older, fully-functioning AutoTrace system requires Matlab and several python dependencies related to GTK+. Because of GTK+, this version is best run on Linux, although it can be installed on Mac OS X using MacPorts (http://www.macports.org/).
##Matlab-free version (under-development)
A Matlab-free version is currently under development. This new version will require PyQt4 to be installed, which is available
from http://www.riverbankcomputing.com/software/pyqt/intro
PyQt4 requires Qt to be installed which can be obtained from http://qt-project.org/
The code for training deep networks is adapted from the author's translational-DBN repository (https://github.com/jjberry/translational-DBN)
which requires gnumpy.py (http://www.cs.toronto.edu/~tijmen/gnumpy.html) to be on the system's Python PATH.
The use of gnumpy.py allows the network to be trained on a CUDA-capable GPU if present, using the cudamat library (https://code.google.com/p/cudamat/)
If no GPU is present, gnumpy.py will use the CPU, which requires npmat.py (http://www.cs.toronto.edu/~ilya/npmat.py).
The Qt-based tools are cross platform, and have been tested on Windows 7, Linux, and Mac OS X, with Qt 4.8.