Multi-modal versatility
Annotate scanned pages, photos of forms, and native digital documents in one place.
Documents
Turn every document into structured, model-ready data.
A professional platform for annotating text, forms, and scanned documents into clean training datasets for OCR and NLP models that extract information without errors and automate document workflows.
Applications
Financial & tax documents
Annotate invoices, tax returns, and receipts so models auto-extract details, amounts, and dates and reduce manual entry and accounting errors.
Technical edge
AI-powered automation
Pre-label recurring fields and layouts so OCR and NLP models learn document structure faster.
Quality control
Reviewers and validation rules catch misread fields and mislabeled regions before the dataset reaches training.
Collaborative workflow
Route large document batches to annotators and reviewers, each with the right permissions.
Customization & integration
Define custom fields, entities, and tags, then export to your OCR or NLP pipeline via API.
Enterprise-grade security
Keep sensitive records confidential, with on-premise deployment for regulated data.
Annotation types
Bounding box
Rectangular boxes that select text blocks, stamps, signatures, or characters for OCR.
Oriented bounding box
Rotated rectangles for text or stamps on angled or skewed scans.
Polygon
Flexible shapes that outline handwritten notes, damaged text, or logos.
Points
Single points that mark checkboxes (OMR) and reference points to verify form completion.
Lines & multilines
Trace underlines, table separators, and handwritten rules in forms.
Skeletons
Connected keypoints that model handwritten characters or graphics like diagrams and charts.
Cuboid
3D boxes for annotating packing slips or documents in 3D space.
3D point cloud
Spatial labeling that assesses the physical condition of documents or archives via 3D scanning.
Semantic segmentation
Pixel-level classification of document regions: text fields, images, tables, and headers.
Instance segmentation
Pixel-wise separation of individual words or characters.
Bitmap
Raster masks that capture fine details like watermarks or paper texture.
Mesh
Polygonal meshes that create digital copies of documents, preserving bends and surface deformation.
Custom
A mix of annotation methods built around your specific document-processing task.
Put your document data to work
Book a demo and we'll scope your document dataset with you.
Custom solution? hello@keylabs.ai
