Automated Document Flow
High workflow efficiency with a minimum number of errors involves storing electronic versions of documents and fast digitization of printed versions of documents. To improve the quality and speed of this process is proposed to carry out automatic recognition of the main image information from a printed document to generate an electronic document template. The solution uses neural network information recognition algorithms from images.
Input Data
- Markup document template
- Supporting fields of the document with their characteristics
- Document Image Set
Output Data
- Character highlighting Model OCR
- Search text blocks Model Text Detection
- Tables recognition Model Table Detection
- Construction of electronic version of document Model
Main Results
The solution is integrated into the business process of the customer:
- Reducing the burden on specialists by 60%
- Increasing the speed of the document in the business process by 10%
- Reduction of costs due to lost documents by 5%
Built solutions for automating recognition:
- Checks and receipts
- Invoice
- Purchase order
- Contracts
Recognition quality:
| Model | Quality |
|---|---|
| OCR | 96 % |
| Text Detection | 98 % |
| Document Classification | 99 % |
| Document Recognition | 94 % |
Example of image Recognition quality: recognition of a printed document
