Smart Document Engine enables local OCR processing on smartphones, allowing businesses to process primary documents like invoices and acts without sending images to the cloud. The system corrects geometric distortions, classifies document types, and extracts business-critical fields such as dates, amounts, VAT rates, and nomenclature. It also processes tables, handwritten notes, checkboxes, barcodes, and signatures.
The on-device architecture significantly reduces security risks by eliminating the need to transfer images to external OCR providers or verification centers. This approach aligns with Russian laws such as 98-FZ on commercial secrets and 149-FZ on information protection, as the original image remains within the organization's controlled perimeter.
The technology supports a wide range of documents, including UPD, UKD, TORG-12, transport waybills, and KS-2/KS-3 forms. Processing a single A4 page takes approximately 2–3 seconds using only the CPU, without requiring GPU or NPU acceleration. This allows data to be integrated into ERP or 1C systems immediately at the point of document receipt.
