OCR Baseline: Benchmark Documents for Invoice, Receipt, and Logistics Processing

A benchmark dataset for optical character recognition and document understanding, created by Timokerr and last updated on 2026-04-22. It contains source PDFs and corresponding JSON ground truth labels for documents across three domains: invoices, receipts, and logistics. The dataset is designed for benchmarking performance across 17 LLM models.

Use Cases

• Benchmarking OCR model accuracy based on ground truth JSON labels.
• Training document information extraction systems based on the invoice, receipt, and logistics domains.
• Evaluating layout understanding capabilities based on the provided PDF and label pairs.

Strengths

• Provides structured ground truth labels in JSON format for each source document.
• Covers three distinct, commercially relevant document domains: invoices, receipts, and logistics.
• Designed for benchmarking against 17 different LLM models, suggesting a standardized evaluation framework.

Limitations

• Description metadata is limited; actual data quality, scale, and column definitions require manual inspection after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.

Provenance

Source

Timokerr on Hugging Face.

Collection Method

Likely collected and annotated for benchmarking purposes.

Time Range

null

Freshness

Last updated 2026-04-22 18:09:36.

Geography

null