Invoices and Receipts OCR V2: 1,000-10,000 Financial Document Images

Containing between 1,000 and 10,000 records of financial documents for OCR tasks, released by user mychen76 in September 2024. It provides paired image and text data for invoices and receipts, formatted as Parquet files for use with the Hugging Face datasets library.

Use Cases

• Training OCR engines to recognize text within image fields
• Fine-tuning document AI models for text extraction from financial layouts
• Benchmarking receipt parsing algorithms on US-based document formats

Strengths

• Contains 1,000 to 10,000 records
• Dual-modality support with both image and text data
• Optimized for performance using Parquet format

Limitations

• Geographic bias toward the United States
• Undocumented column schema
• Unknown license status

Provenance

Source

mychen76 via Hugging Face

Freshness

Last updated September 2024.

Geography

United States