DENKnet OCR - robust character recognition without mixing up characters
OCR results now recognise relationships between individual characters using new output-filters
Transforming printed or handwritten text from camera images into machine-readable text data is still one of the most challenging tasks for OCR systems. Recognizing very similar characters separates the wheat from the chaff. Mixing up characters such as "O" and "0" or a "4" that is recognized as an "A" can hardly be ruled out, even with very good input data.
The Deep OCR model of DENKnet OCR is prepared for such extreme cases. Besides recognizing individual characters, it now also takes their relationships into account with extremely high reliability. This is made possible by new output filters that allow to define more precisely the format of a result word that is to be displayed only. For example, when reading out dates, a date format (e.g. "01/01/0123") can be specified.
See for yourself how reliably and quickly optical character recognition can work. Test the new DENKnet OCR or simply visit us at VISION in October - where the technology will be used in one of our demo systems.
