Deep OCR

Deep OCR is a deep learning-based method for OCR. This new technology brings machine vision one step closer to human-like reading.

In comparison to existing algorithms, Deep OCR can identify characters regardless of their orientation and font type. Its ability for automatic character grouping enables the identification of entire words, significantly enhancing detection performance as it can prevent misinterpretation of characters with similar appearances.

Training Deep OCR

MVTec’s Deep OCR empowers users to efficiently tackle text reading in many use cases. Starting from HALCON 22.05, this technology has been enhanced with trainable capabilities. The possibility to train specific applications on your dataset allows you to tackle even the most complex scenarios like reading text with poor contrast (e.g., on tires). Another advantage is the ability to train very rare characters or font styles. Training for Deep OCR significantly improves performance and usability, making applications more robust.