unstructured-inference 0.5.1

A library for performing inference using trained models.

Open-Source Pre-Processing Tools for Unstructured Data

The unstructured-inference repo contains hosted model inference code for layout parsing models. These models are invoked via API as part of the partitioning bricks in the unstructured package.

Installation

Package

Run pip install unstructured-inference.

Detectron2

Detectron2 is required for using models from the layoutparser model zoo but is not automatically installed with this package. For MacOS and Linux, build from source with:

pip install 'git+https://github.com/facebookresearch/detectron2.git@e2ce8dc#egg=detectron2'

Other install options can be found in the Detectron2 installation guide.

Windows is not officially supported by Detectron2, but some users are able to install it anyway. See discussion here for tips on installing Detectron2 on Windows.

PaddleOCR

PaddleOCR is required for table processing for x86_64 architectures. It should not be installed under MacOS with Apple Silicon cpu.

PaddleOCR should be installed using the following instructions.

pip install "unstructured.PaddleOCR"

Repository

To install the repository for development, clone the repo and run make install to install dependencies. Run make help for a full list of install options.

Getting Started

To get started with the layout parsing model, use the following commands:

from unstructured_inference.inference.layout import DocumentLayout

layout = DocumentLayout.from_file("sample-docs/loremipsum.pdf")

print(layout.pages[0].elements)

Once the model has detected the layout and OCR'd the document, the text extracted from the first page of the sample document will be displayed. You can convert a given element to a dict by running the .to_dict() method.

Models

The inference pipeline operates by finding text elements in a document page using a detection model, then extracting the contents of the elements using direct extraction (if available), OCR, and optionally table inference models.

We offer several detection models including Detectron2 and YOLOX.

Using a non-default model

When doing inference, an alternate model can be used by passing the model object to the ingestion method via the model parameter. The get_model function can be used to construct one of our out-of-the-box models from a keyword, e.g.:

from unstructured_inference.models.base import get_model
from unstructured_inference.inference.layout import DocumentLayout

model = get_model("yolox")
layout = DocumentLayout.from_file("sample-docs/layout-parser-paper.pdf", model=model)

Using models from the layoutparser model zoo

The UnstructuredDetectronModel class in unstructured_inference.modelts.detectron2 uses the faster_rcnn_R_50_FPN_3x model pretrained on DocLayNet, but by using different construction parameters, any model in the layoutparser model zoo can be used. UnstructuredDetectronModel is a light wrapper around the layoutparser Detectron2LayoutModel object, and accepts the same arguments. See layoutparser documentation for details.

Using your own model

Any detection model can be used for in the unstructured_inference pipeline by wrapping the model in the UnstructuredObjectDetectionModel class. To integrate with the DocumentLayout class, a subclass of UnstructuredObjectDetectionModel must have a predict method that accepts a PIL.Image.Image and returns a list of LayoutElements, and an initialize method, which loads the model and prepares it for inference.

Security Policy

See our security policy for information on how to report security vulnerabilities.

Learn more

Section	Description
Unstructured Community Github	Information about Unstructured.io community projects
Unstructured Github	Unstructured.io open source repositories
Company Website	Unstructured.io product and company info