Optical Character Recognition
Use artificial neural networks to locate text and characters
in images and extract said text.
Various Font Learning
The typefaces used to create the document vary greatly in the languages of each country. At first, the most used fonts are already pre-learned, and it is possible to learn additional fonts to meet other purposes as well.
Quadrilateral Region Detection
Many text-carrying mediums, such as forms, cards, cheque, business cards, and books are quadrilateral in shape. By extracting and flattening quadrilateral/rectangular regions from images, we are able to analyze and extract text more accurately. This increases the chance of retrieving textual information from images that are taken with mobile devices, where the camera angle unintentionally projects a perspective to the object in question.
Extract Text and Character
The most important starting point for increasing the accuracy and recognition rate of OCR is to pinpoint where texts and characters are in the document. Artificial neural networks accurately extract the position of text or characters within an image.
Corresponds to Various Image Quality
Corresponds to the quality of the original image, depending on the quality of the original document and the quality of shooting or scanning. It supports character recognition by performing an image pre-processing algorithm optimized for article identification and text extraction.
Light Modules and Swift Process
Lightweight modules and fast processing speeds are useful for extracting text from large volumes of documents, which can be applied and leveraged in mobile applications.
Field of Usage Examples
Document text recognition
and indirect identity authentification (e-KYC)
Automatic business card recognition/organization ID card classification