← A PassionLabs product
PageSense / Document AI
Make sense of every page.
From scanned pages to structured data. PageSense detects document regions and extracts text from PDF and image inputs, with annotated results for human inspection.
document_analysis.pdf
What it does today
From pages to structured results.
Focused on extraction, transparency, and making the output easy to inspect.
01 / INPUT
PDFs & images
Accept page images and multi-page PDFs, rendering PDF pages for analysis.
02 / ANALYZE
Layout + OCR
Use DocLayout-YOLO and PaddleOCR to identify regions and recognize text.
03 / REVIEW
JSON + annotated PDF
Export machine-readable regions and a visual PDF for inspecting detections.
How it works
An inspectable processing pipeline.
STEP 01
Upload locally
Start with PDFs or supported image files in the development pipeline.
STEP 02
Detect & read
Render, find regions, clean overlaps, and recognize text in reading order.
STEP 03
Inspect outputs
Review the extracted JSON and annotated PDF; correct OCR issues when necessary.
Development preview
Built experimentally. Tested carefully.
PageSense is not accepting public document uploads. The local pipeline handles PDF/image analysis, and a single PDF was successfully tested in a CPU-only Docker experiment. Public APIs, robust question reconstruction, and reliability benchmarks remain future work. Please don't submit sensitive documents to unofficial demos.