Document OCR Function Description:

Document OCR (Optical Character Recognition) focuses on "accurately extracting text from images and efficiently converting it into editable content." 

It supports multiple document types and various scenario requirements, enabling text digitization without manual input. Core functions include:

Multi-type document adaptation and recognition: Processes scanned copies of paper documents, electronic images (JPG/PNG, etc.), PDF files (including scanned versions and image-layer PDFs), 

handwritten notes, and various certificates (ID cards, business licenses, invoices, bank cards), covering common office, daily life, and government 

document scenarios. Direct recognition is possible without format conversion.

High-precision multilingual recognition: Supports recognition of multiple mainstream languages including Chinese, English, Japanese, Korean, French, and German. Accurately recognizes 

blurred text (e.g., unclear printing, faded text in old documents), tilted documents (automatic correction from 0-30° tilt), and handwriting (95% 

accuracy for neat handwriting), reducing errors. It also distinguishes text formatting (e.g., bold, italic, underline), preserving the original layout logic.

Structured Content Extraction:

Beyond extracting plain text, it automatically categorizes fields by document type—for example, when recognizing invoices, it accurately extracts 

"invoice number, invoice date, buyer information, amount, and tax"; when recognizing contract documents, it extracts "names of both parties, 

contract validity period, and key terms and keywords," automatically generating structured tables (Excel)  or annotated documents, eliminating 

the need for manual field organization.

Convenient Proofreading and Optimization: 

After recognition, a comparison preview page of "original image + recognized text" is generated. Users can directly annotate erroneous text on 

the page (e.g., areas of recognition deviation), and the AI will automatically learn and optimize based on the annotations. It also supports 

"keyword search," allowing quick location of specific content in the recognized text (e.g., searching for "payment period" across multiple contracts), 

improving subsequent review efficiency.

Image

Text extraction results:

公司简介

深圳市壹贝斯科技有限公司成立于2021年，专注于网络通讯和AI应用领域,为客户提供面向各垂直行业的解决方案和ODM服务, 产品覆盖网络通讯, 边缘计算, 白盒交换机, 汇聚分流, 边缘AI等设备; 以及各类CPU核心板,包括Type6, Type7 COMe及高性能COM-HPC计算模块; AI模块包括Hailo AI 单芯片miniPCIe & M.2 卡以及多芯片PCIe & MXM卡; Nvidia Jetson系列模组和板卡定制服务以及Nvidia Quadro GPU MXM卡, 包括Turing/Ampere/Ada Lovelace 系列。