samarth010/nano-receipts

🧾 Nano Receipts Dataset

A diverse collection of 2428 hyper-realistic synthetic receipt images generated using state-of-the-art text-to-image AI models.

🚀 Quick Start

from datasets import load_dataset

# Load dataset (fast parquet format!)
dataset = load_dataset("34data/nano-receipts")

# Access images
image = dataset["train"][0]["image"]  # PIL Image
filename = dataset["train"][0]["filename"]

📊 Dataset Details

• —Total Images: 2428 receipts
• —Format: Parquet (optimized for fast loading)
• —Image Format: PNG
• —Generation: Text-to-image synthesis (Google Gemini 2.5 Flash Image)
• —Quality: Hyper-realistic synthetic receipts

🏪 Receipt Types (28+ Categories)

Physical Receipts (Thermal Printed)

• —🛒 Retail (Walmart, Target, CVS, Walgreens, etc.)
• —🍽️ Restaurants (casual & fine dining)
• —🍔 Fast Food (McDonald's, Burger King, Taco Bell, etc.)
• —☕ Coffee Shops (Starbucks, Dunkin', Peet's, etc.)
• —⛽ Gas Stations (Shell, Chevron, Exxon, etc.)
• —🅿️ Parking receipts
• —💊 Pharmacies
• —✂️ Salons & beauty services
• —🚗 Car washes

Digital Receipts & Confirmations

• —🚕 Rideshare (Uber, Lyft)
• —🍕 Food Delivery (DoorDash, Uber Eats, Grubhub)
• —📦 E-commerce (Amazon, eBay, Etsy, Walmart.com)
• —📺 Subscriptions (Netflix, Spotify, Adobe, etc.)
• —💸 Money Transfers (Venmo, PayPal, Zelle, Cash App)
• —🛣️ Toll Roads (E-ZPass, FasTrak, SunPass)

Service Invoices

• —🏥 Medical billing statements
• —💡 Utility bills (electric, gas, water)
• —🏨 Hotel folios
• —✈️ Airline e-tickets
• —🔧 Car service & repairs
• —🏠 Home services (plumbing, electrical, cleaning)
• —🚙 Car rentals (Hertz, Enterprise, Budget, Avis)
• —🐾 Pet services (grooming, vet, boarding)
• —💪 Gym memberships
• —🎬 Entertainment tickets

Financial Documents

• —🏧 ATM transaction receipts
• —📮 Shipping receipts (USPS, FedEx, UPS, DHL)

🎯 Use Cases

• —Receipt OCR Training: Train and test OCR models on diverse receipt formats
• —Document Classification: Multi-class receipt type classification
• —Layout Analysis: Study receipt structure and layout patterns
• —Expense Tracking AI: Build automated expense categorization systems
• —Financial Document Understanding: Train models to understand financial documents
• —Fraud Detection: Develop receipt authenticity verification systems
• —Mobile Banking Apps: Test receipt scanning features
• —Accounting Software: Receipt parsing and data extraction

📁 Dataset Structure

Dataset({
    features: ['image', 'filename', 'receipt_id'],
    num_rows: 2428
})

• —image: PIL Image object (PNG format)
• —filename: Original filename (e.g., receipt_t2i_000123.png)
• —receipt_id: Sequential ID extracted from filename

⚡ Why Parquet Format?

This dataset uses Parquet format for optimal performance:

• —10-100x faster loading compared to individual image files
• —Efficient storage with automatic compression
• —Streaming support for large datasets
• —Direct integration with HF datasets library
• —Zero-copy reads for maximum speed

🔧 Generation Process

Each receipt was generated with detailed prompts specifying:

• —Business/service name and realistic locations
• —Transaction dates and times
• —Itemized purchases or services
• —Realistic pricing, taxes, and totals
• —Payment methods (cash, credit card, digital)
• —Authentic formatting for each receipt type
• —Format-appropriate rendering (thermal, digital, printed)

📝 License

MIT License - Free to use for commercial and non-commercial purposes.

🔄 Updates

This dataset is actively maintained. To add more receipts:

# Generate more receipts
python3 generate_receipts_t2i.py

# Upload incrementally
python3 upload_to_hf_parquet.py

📚 Citation

@dataset{nano_receipts_2024,
  title={Nano Receipts: Synthetic Receipt Dataset},
  author={34data},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/34data/nano-receipts}
}

🤝 Contributing

Found an issue or want to contribute? Open an issue on the repository.

Generated with ❤️ using AI text-to-image models