AI Documents & PDF Tools
PRODUCT OVERVIEW
OCI Document Understanding is an Oracle Cloud Infrastructure AI service for developers who need to turn document files into structured data. Using optical character recognition and other pretrained models, it extracts text, tables, and key-value pairs from scanned documents, mobile uploads, PDFs, invoices, receipts, passports, driver's licenses, and health insurance ID cards, and classifies documents into categories such as invoice, receipt, or résumé. Pretrained OCR and key-value models support multiple languages. Teams can also train, evaluate, deploy, and analyze custom models with their own data through the OCI console, subject to limited availability. The service is called through REST APIs, Python and Java SDKs, or the OCI command line, and results can be stored as JSON in OCI Object Storage and integrated with other OCI services. Pricing is pay-as-you-go with free monthly transaction allowances and per-unit charges beyond them.
Choose a file from an OCI Object Storage input bucket or upload a local file from your computer.
Use the pretrained models to extract text, tables, key-value pairs, and document classifications automatically.
Upload, label, and train a custom model in the OCI console to tailor extraction to your data or use case; this capability has limited availability.
Save results in JSON format to an OCI Object Storage output bucket and integrate with other OCI services such as Oracle APEX and OCI Digital Assistant.
Automating accounts payable and expense processing by extracting invoice and receipt data
Document analytics: extracting document information into Oracle Autonomous Data Warehouse and visualizing it in Oracle Analytics Cloud
Document process automation: prefilling fields in web applications or triggering business processes and human review steps in OCI Process Automation
Enhanced search and retrieval through document classification
Customer onboarding with extraction from identity documents such as passports and driver's licenses
It processes scanned documents, mobile uploads, and PDFs, including pages that are rotated, tilted, or shaded. It also handles invoices, receipts, passports, driver's licenses, and health insurance ID cards.
Pretrained OCR and key-value pair models support Arabic, Chinese, Dutch, English, French, German, Hebrew, Japanese, Portuguese, Russian, Spanish, and Ukrainian.
Through REST APIs, multiple SDKs including Python and Java, or the OCI command line.
Yes. You can upload, label, and train custom models for key-value pairs and document classification in the OCI console, then evaluate, deploy, and analyze them. This capability has limited availability.
Results are stored in JSON format in an OCI Object Storage output bucket and can be integrated with other OCI services such as Oracle APEX and OCI Digital Assistant.
Pricing is pay-as-you-go per 1,000 transactions or per training hour. Each line item includes a free monthly allowance of the first 5,000 transactions, or the first 15 training hours for custom training, with charges beyond that.
Oracle states that the models do not store any data for training, debugging, or other purposes.
Oracle offers a free trial and a workshop with demo documents and your own documents.
Pay-as-you-go pricing in USD. Each Document Understanding line item includes a free monthly allowance of the first 5,000 transactions (or first 15 training hours for custom training); charges apply only beyond those allowances. A free trial is available.
Optical character recognition for extracting text from document files. The first 5,000 transactions per month are free.
Extraction of document properties using pretrained models. The first 5,000 transactions per month are free.
Extraction of key data from documents using pretrained models. The first 5,000 transactions per month are free.
Document property extraction with custom-trained models. The first 5,000 transactions per month are free.
Key-value extraction with custom-trained models. The first 5,000 transactions per month are free.
Training of custom models with your own data. The first 15 training hours are free.
COMMUNITY REVIEWS
No reviews yet.
