Core Toolkit

The Core Toolkit is a collection of standard building blocks (written in C++) and build tools that are necessary to build online handwriting recognizers for new scripts for Windows or Linux.  The Core Toolkit is targeted at researchers, and developers with some working knowledge of building recognition systems. It is also designed to be extended by adding new features, shape recognizers, and so on. The core toolkit currently primarily supports isolated shape/character recognition, and recognition of boxed fields of text.

Core Toolkit comes bundled with Lipi Toolkit.

The Core Toolkit includes:

• Generic Class and Utility Libraries: Common classes such as Trace, TraceGroup, reading/writing UNIPEN etc
• Common preprocessing: Size normalization, resampling, etc
• Features: PointFloat, npen, etc
• Shape classifiers: k-Nearest Neighbor, Active-DTW, Neural Network, ...
• Boxed-field recognizer: For recognition of boxed fields of characters
• Lipi Engine: Main shape and word recognition interface for client applications
• Tools - Training and testing, viewing model data, evaluation of accuracy and errors
• Makefiles and utility scripts - Building and packaging recognizers
• Sample programs: Test application source code to create recognizers and invoke recognition functions
• Sample handwriting data for testing

The Core Toolkit can be built on both Windows and Linux for deployment on those platforms.

Detailed descriptions of the components and usage of Core Toolkit are available in the User Manual and the Reference Manual (out of date - to be updated !).