Azure Invoice Recognizer
Applies advanced machine learning to accurately extract from forms and tables in documents
The invoice model combines powerful Optical Character Recognition (OCR) capabilities with deep learning models to analyze and extract key fields and line items from sales invoices. Invoices can be of various formats and quality including phone-captured images, scanned documents, and digital PDFs. The API analyzes invoice text; extracts key information such as customer name, billing address, due date, and amount due; and returns a structured JSON data representation.
You can list a component in the marketplace, and define if you want it to be a template.
✅ Available in Marketplace
❌ Can not be used as a template
What is a Model?
This component is a model, which is a type of store that is specialized for handling AI/ML model storage, which includes both the implementation, and the results of training.
Models are a foundational part of Kodexa, and are used in many different ways. For example, a model can be used to classify documents, or to extract data from documents. Models can also be used to train other models.
Metadata
✅ Atomic Deployment (Recommended)
❌ Not trainable
Model Runtime
A model needs to reference a model runtime to use.
✅ kodexa/base-model-runtime
The model also has the following model runtime parameters configured. This influences how the model is run, see the model runtime references to determine what parameters are available.
| Parameter Name | Value |
|---|---|
| module | azure_models |
| function | invoice_infer |
Inference Options
When you use the model for inference, you can use the following options:
| Option Name | Default | Required? | Type | Description |
|---|---|---|---|---|
| store_azure_output | False | None | boolean | Store the JSON response from Azure in the document as a feature on the root node |
| keep_azure_lines | True | None | boolean | Retain the line layout from Azure in the document |
Model Label Taxonomy
This model provides a taxonomy of labels that can be applied, these labels are not used to extract data but are "meta-labels" that are in place to help the model train on the content.