Computer Vision · Finance
Danish Invoice AI Processing System
A production system reading Danish invoices from unstructured PDFs — computer vision identifies layout regions, NLP classifies entities semantically, and structured output flows straight into ERP workflows.
YOLOv8BERT NERTesseract OCRFlaskPython
90%
Reduction in manual entry
95%
mAP region detection accuracy
What we built
01
End-to-end CV + NLP
A full pipeline built specifically for Danish-language document structure.
02
Region detection
YOLOv8 fine-tuned to 95% mAP accuracy on invoice layout regions.
03
Semantic classification
BERT-based NER resolving Fakturanummer, Beløb, Leverandør and Udløbsdato.
04
ERP integration
Deployed as a high-performance Flask REST API with direct ERP compatibility.
Next case study
AI-Driven Healthcare Copay Management