cff-version: 1.2.0
message: "If you use this benchmark, please cite it as below."
type: dataset
title: "Invoice Extraction Benchmark v1: 181 synthetic invoices, photos, scans and handwritten documents with answer keys"
abstract: >-
  A synthetic test set for invoice data extraction: 181 documents (clean PDFs,
  phone photos and scans at three damage grades, handwritten invoice books,
  credit notes, dual-currency and VAT-inclusive invoices, multi-page invoices,
  several invoices in one file, supplier statements, and invoices whose printed
  totals deliberately do not add up), with an answer key per document, a scorer,
  and the generator that reproduces the documents from seed 42. Made by
  InvoiceParser Pro, an invoice extraction vendor.
authors:
  - name: "InvoiceParser Pro"
    website: "https://invoiceparserpro.com"
version: "1.0"
date-released: "2026-10-04"
repository-code: "https://github.com/DrewKraken/invoice-extraction-benchmark"
url: "https://invoiceparserpro.com/accuracy"
license: CC-BY-4.0
keywords:
  - invoice extraction
  - invoice OCR
  - document AI
  - intelligent document processing
  - accounts payable
  - benchmark
  - synthetic dataset