Within the range of the classification as well as the recognition of texts (Optical Character Recognition, OCR) the software of MVTec offers technologies, which are based partially and also holistically on Deep Learning. These enable ideal results even under difficult conditions, such as the identification of slanted text, distorted letters or characters printed or etched on reflective surfaces or strongly structured color backgrounds.

Deep OCR

HALCONs Deep OCR, a holistic deep-learning-based approach for OCR, localizes characters in a very robust way, even regardless of their orientation and font type. The ability to automatically group characters allows the identification of whole words which strongly increases the recognition performance as, e.g., misinterpretation of characters with similar appearances can be avoided.

Traditional OCR

With many pretrained fonts (some based on deep learning technology) included from a wide range of industries (dot matrix fonts, semi font, industrial fonts, hand written fonts, etc), customers can achieve excellent recognition rates “out of the box”. MVTec’s OCR tools scored a 0.65% error rate on the MNIST data set. MVTec’s software also provides time-saving tools and assistants for training custom fonts.
Another critical aspect of OCR, at which MVTec excels, is segmentation of characters from the image background. MVTec’s unique arbitrary regions of interest capability combined with their leading blob analysis tools and comprehensive image filtering techniques, allows developers to effectively isolate and extract characters from complex backgrounds resulting in more accurate character classification and better reading rates. Our software reads a single character in less than 0.1 ms.

Application examples

Find out how customers use MVTec software to develop high-innovative solutions with OCR and OCV.

Learn more