For all users of the Cubert Utils software environment we are now able to offer the full integration of online-classification capability. With the new perClass plug-in users will experience a seamless integration of hyperspectral applications into our camera platform.
Advanced machine learning technology on hyperspectral images
Cubert GmbH and PR Sys Design are happy to bring advanced machine learning technology to Cubert ecosystem. perClass software package, developed by PR Sys Design, enables researchers and industrial R&D specialists to create powerful classification solutions. Main contribution of this cooperation is that Cubert users can, within five minutes, record spectral images, create statistical classifiers for specific materials and deploy such solution processing a live data stream. This enables the users to identify types of plastics, sort natural products such as vegetable or nuts, classify types of tissues and biological material, analyze tree types and plant health in remote sensing images and more.
Full integration of classification engine reduces time to market drastically
perClass comes tightly integrated within the 2.0.4 “Fuchsia” release of Cubert Utilities. Users may install perClass Toolbox for Matlab(R) directly via the Cubert installer. Multi- and hyperspectral images, acquired by the complete range of Cubert cameras, may be directly imported into perClass environment. A rich set of interactive visualization, data modeling and classification tools are available for rapid prototyping. Unique concept of perClass pipelines allows users to quickly model and test diverse data processing chains. This includes practical setups such as one-class detectors, multi-class discriminants and cascaded systems. Classification algorithms can process entire spectrum or only specific, automatically selected, spectral bands. Solution performance can be fine-tuned based on application-specific requirements, mitigating specific types of errors. The major improvement over other machine learning toolkits is the direct deployment of any data processing pipeline outside perClass environment. Classification models can be directly exported as Cubert plugins and applied to pre-recorded or live data in Cubert Touch application. perClass integration into Cubert ecosystem significantly lowers time to market for a new generation of intelligent products based on advanced machine learning technology.