InstantVision Toolboxes for MATLAB®
A set of toolboxes has been developed to design, develop,
implement and test complex algorithmic solutions within Matlab
environment. The set includes a simulation environment for specific
cellular visual microprocessor functions (MatCNN) and connected
system level packages to build multi-target tracking
(MatTrack) and feature classification (MatClass) algorithms based
on the InstantVision
ISE libraries.
MatCNN - AnaLogic CNN Simulation Toolbox for MATLAB
An easy to use toolbox for designing single-layer analogic
celullar neural/nonlinear network (CNN) based algorithms with
nearest and second neighbor linear and nonlinear interactions. The
toolbox supports the simulation of CNN based Cellular Visual
Microprocessor (CVM) cores with various cell, synaptic
and output characteristics and array properties.
MatClass - Feature Classifier Toolbox for MATLAB
A feature classifier toolbox based on the FC Library of the
InstantVision ISE with expanded functionality and tools for
off-line design, analysis and testing. The toolbox contains the
implementation of a set of classifiers (e.g. ART and NN based
classifiers), various off-line feature selection and optimization
methods (e.g. PCA based) with video marking and generating tools in
order to support a complex classifier system
design.
MatTrack - Multi-target Tracking Toolbox for MATLAB
A multi-target tracking toolbox based on the MTT Library of
the InstantVision ISE with expanded functionality and tools for
off-line design, analysis and testing. The toolbox contains
the implementation of distance calculation methods (e.g.
city-block based), data assignment and association strategies
(e.g. ENN and JVC), state prediction filters (e.g. IMM) with
video marking and debugging tools in order to support a
complex multi-target tracking system design.
Note: all toolboxes have
been developed within academic cooperation with PPCU and SZTAKI and are provided
free of charge for non-commercial use only. No technical
support is provided yet for these packages. Comments could be sent
to this email address. |
