Samsung SDS image pre-processing technologies and text recognition algorithms help eliminate any distractions in images such as rotation, noise, and watermark, guaranteeing the highest text recognition accuracy in the industry. Accurate character recognition is available in 97 languages irrespective of the type and quality of documents.
A deep learning-based image analysis model with a digital image processing algorithm provides findings after identifying sentence structures such as sentences, tables, images, and non-character columns and distinguishing tables from key-value.
AICR provides analysis for not only general documents but also receipts and logistics documents (invoices, bills of landing, etc.). Clearly distinguishing sentences, tables, and key values from invoices and extracting data promote efficiency in financial tasks.
Data can easily be integrated with the application using the AICR-extracted data through API calls. Uploading target images for analysis and storing findings are done easily using Object Storage on the Samsung Cloud Platform.
- Types of Image : Document, invoice
- Types of extracted data
· Text : Extract text and location
· Table : Extract text and location of each cell by distinguishing table structure
· Format : Extract data in key-value by distinguishing document format
- Supported languages for character recognition : 7 language family, 97 languages (Korean, 56 in Latin, 11 in Cyrillic, 25 in Arabic, 2 in Chinese, Japanese, and Thai)
- AICR functional/performance testing through Demo
- Demo limitations : Use JPG, PNG, PDF format, no bigger than 10MB, one page
- Image format : JPEG, PNG, PDF
- Maximum capacity and page limit
· Console demo and synchronous API : 10MB, 1 page
· Asynchronous API : Image file (JPEG, PNG) 10MB, PDF file 50MB/150 pages
- View call status by API and by day/week/month
- View the number of calls by API (All/success/error)
- View response time for successful API