ORIGIEN Next-Gen Digital Infrastructure
Distributed Core Banking System
Distributed Financial PaaS
Data Middle Platform
Digital Marketing
Risk & Compliance
Payment & Settlement
Channel Solutions
Credit Management
Transaction Banking
Open Banking
Digital Operation
QA Platform
AI Applications
Industry Solutions
GienTech's Intelligent ICR solution solves character recognition for various financial image documents using proprietary AI models. The AI+OCR approach ensures high-precision text recognition, high-quality content validation, and accurate structured output, supporting scenarios like data entry, ID recognition, document and contract identification/comparison.
Document Classification | Quickly classifies and archives uploaded documents and processes them structurally based on classification results.
Image Preprocessing | Uses techniques such as stamp detection/removal, angle correction, and denoising to preprocess images affected by skew, folds, or stamp interference.
Text Detection | Uses industry-specific pre-trained models to accurately detect valid text positions.
Layout Analysis | Analyzes document layouts to assist in recognition, greatly improving structured recognition accuracy.
Text Recognition | Recognizes over 400 common printed and handwritten fonts, supporting both Chinese and English mixed texts.
Structured Output | Analyzes and processes various formats of contracts and vouchers to achieve optimal structured recognition accuracy.
This solution leverages AI technology to build four core models:
-
ICR Recognition Model
A proprietary model developed through deep customization using machine learning algorithms and open-source frameworks. It is trained with large volumes of text and real document samples to ensure high-speed and high-accuracy recognition. The model also supports secondary development and retraining.
-
Text Detection Model
Designed to identify text within document images—such as skewed or distorted characters—this model employs deep learning-based techniques for accurate text detection. The algorithm and training data for text localization are entirely proprietary, and the model supports secondary development and retraining.
-
Seal Detection/Recognition/Removal Model
In financial documents, seals often obscure key content. This model detects seals, recognizes circular text within them, and removes the seal to prevent interference with content recognition underneath.
-
Binarization Model
This model automatically generates binarization thresholds, converting image pixel grayscale values to binary (0 and 1). This enhances image clarity for optimal text recognition and detection accuracy.
-
ICR Recognition Capabilities
The ICR engine delivers high recognition accuracy, with strong text localization and structured output capabilities. It supports image pre-processing to address common issues in the financial sector, such as blurred images and watermark interference. The model can recognize both printed Chinese and English text, partially recognize handwritten Chinese, and identify seal content.
-
Product Maturity
The solution offers both SDK and API access, and supports standalone deployment of the full ICR system. A comprehensive application is also available, featuring built-in recognition functions for various types of bills and documents.
-
Coverage of Document Types
Out-of-the-box support is provided for over 20 common types of documents and certificates. Given the high degree of customization in document formats, tailored training services are also available to meet specific requirements.
-
GienBot RPA
鲸Bot RPA作为数字化劳动力产品解决方案，基于界面元素识别和视觉反馈技术，融合模法师机器学习平台、鲸图知识图谱平台能力，实现企业业务流程自动化，提升企业经营管理的水平
Experience the Product -
ORIGIEN AI Development and Service Platform
The ORIGIEN AI Development and Service Platform is an efficient, low-threshold, open and secure one-stop enterprise-level AI middle platform. Using industry-leading technologies, it builds a full-link, end-to-end AI algorithm development suite and tools. It integrates complex model research and development with service delivery, supporting full-stack machine learning, deep learning, and lifecycle management of large model training and reasoning. The platform helps enterprises achieve the goal of "managing computational power, building models, and utilizing services effectively."
Experience the Product -
AI Training Companion System
Creates realistic business scenario simulations for human-machine interaction training, adopting a closed-loop "Learn-Practice-Test-Evaluate" model to ensure progressive mastery, enabling trainees to achieve true competency and effective communication, while enhancing training management through data-driven operations.
Experience the Product -
Intelligent Dual-Recording System
Delivers an integrated intelligent audio/video recording solution with standardized compliance.
Experience the Product
