The Layout Analysis Service is a high-performance system built on the Swiss AI Center Core Engine, designed for analyzing the layout of image-based documents. It takes an image-based document as input and provides:
A JSON file containing:
The identified components/parts of the document.
Their bounding boxes (bbox).
An annotated image:
Visualizing the bounding boxes.
Displaying the names of the identified parts.
This service is inspired by state-of-the-art research and optimized for real-time performance on mobile and CPU devices, leveraging PaddlePaddle's innovative detection architecture.
PP-PicoDet: A better real-time object detector on mobile devices
Check the related documentation for more information.