PRSD Studio

PRSD Studio consists of two parts:

Matlab toolbox simplifying design of pattern recognition algorithms.
C-based library bringing fast execution of trained classifiers to custom applications without any external dependencies.

System Requirements

PRSD Studio requires Matlab 7.x.

Supported platforms: MS Windows 32-bit/64-bit, Mac OS X Intel, GNU/Linux x86 32-bit/64-bit

Feature Highlights

Interactive data visualization

Work with your data sets using interactive scatter plot. Define multiple labeling schemes (classes, objects, patients, etc). Paint interactively new classes or clusters such as outliers or modes. Hovering over data samples, visualize the original measurements such as images or spectra.

Fine-tune classifiers using ROC analysis

Optimize performance of trained classifiers to application-specific requirements using ROC analysis. Understand the classifier behaviour at different operating points using interactive ROC plots. Use two-class or multi-class ROC analysis leveraging output thresholding, class output weighting, with optional reject options. Optimize algorithms with multiple operating points such as detector/classifier cascades.

Quickly deploy your solutions

Shorten your development time as the transfer of classifiers from design prototype to production application is matter of seconds. The execution library libPRSD provides a simple general interface to your applications. You don't need to specify which algorithms you want to run when embedding the library. Once your application is linked with libPRSD, you can export your trained classifier into a binary file, called pipeline. Your application loads the pipeline file and may directly execute the classifier on new data.