VNPT SmartReader - AI application for digitization and information extraction

Engine OCR

Item	Specifications
Character recognition capability, used for reading Vietnamese printed characters in Vietnamese documents and papers in scan (.pdf) or image (.jpg, .png) formats	Accuracy reaches 99%
Input document standard	Pure text document (printed), no handwriting. Documents containing tables must ensure full rows, columns, and grid lines, and the table must fit within a single page. PDF/image data files must not be blurred, tilted, missing corners, overexposed, or taken indirectly from another device. PDF/image files must not be resized or compressed. Document types provided for OCR must comply with current legal regulations.
Sample training	Capable of training new samples for the OCR engine