Receipt Images with Structured JSON Labels for OCR Fine-Tuning

230 receipt images are labeled with structured JSON data extracted via the Gemini model, covering a variety of merchants, formats, and receipt layouts. The dataset is intended for fine-tuning vision-language models on document OCR tasks and was uploaded by author docjay131 to Hugging Face. It was last updated on April 5, 2026.

Use Cases

• Fine-tuning OCR models based on labeled receipt images and structured text.
• Training document layout understanding models based on the variety of receipt formats mentioned.
• Benchmarking structured information extraction performance from receipts.
• Developing multi-modal models that process both image and JSON text data from documents.

Strengths

• Contains 230 labeled receipt images, providing a base for model training.
• Labels include structured JSON data, which likely contains key-value pairs for receipt fields.
• Covers a variety of merchant types and receipt layouts as noted in the description.

Limitations

• Row count for the JSON data is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's source and collection methodology are not detailed, making bias assessment difficult.

Provenance

Source

huggingface

Collection Method

Extracted via Gemini model, according to the description.

Freshness

Last updated 2026-04-05 05:25:54; freshness should be verified.