API

Otary is built upon three core modules Image, Geometry and Vision, each designed with a distinct architectural pattern to provide flexibility and power.

Image: A Composition-Based Approach

The image module is designed following a composition over inheritance principle. This allows you to build complex image processing pipelines by combining smaller, independent, and reusable components. Instead of inheriting properties from a monolithic class, you can dynamically assemble functionality, leading to a more flexible and maintainable codebase.

The image object is composed of a Reader to load data, Writer to save data, a Transformer to apply modifications and a Drawer to add overlays, all working together seamlessly.

Geometry: An Inheritance-Based Structure

The geometry module uses a more traditional inheritance model. It provides a clear and hierarchical structure for geometric entities. Base classes define common behaviors, and specialized subclasses inherit and extend this functionality. This approach is ideal for creating a well-defined and logical classification of shapes and geometric objects.

Vision: Computer Vision Utilities

The vision module provides tools for manipulating OCR outputs. Otary does not include any OCR engine, but it provides an interface for working with OCR outputs.

You can use this module to process and analyze OCR outputs extracted from your favorite OCR engines, such as Tesseract, EasyOCR, DocTR, Azure Document Intelligence, Amazon Textract, etc...

This way you can easily display OCR outputs on top of your images, analyze your OCR outputs, extract information using a heuristic Key Information or key-value pairs retrieval method, and much more.