Dataset · katanaml-org
Labeled invoices for the Donut model
katanaml-org/invoices-donut-data-v1
Small set of 500 English electronic invoices with ready annotations for extracting structured data from documents.
What's inside
The dataset includes 500 electronic invoice images in English. The Katana ML team annotated them and prepared them specifically for fine-tuning the Donut model, based on the original dataset from Mendeley Data.
What you can do with it
- Fine-tune the Donut model for automatic parsing of invoices and receipts
- Test OCR pipelines and key field extraction systems from financial documents
- Tasks
- Embeddings
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 1.2K
- Likes
- 45
Models trained on it
- katanaml-org/invoices-donut-model-v1Image → text 153
- zackyalgiffari/qwen3.5-2b-receipt-invoice-extract-loraImage text → text 29
- PCS/Extract-O-MaticImage → text 13
- PCS/Extract_MaticImage → text 7
- MH20011981/donut_trained_invoices-donut-data-v1Image text → text 7
- ssraut/Extract_MaticImage → text 6
- rationalenterprise/fine_tuned_image_relevance_model 0
How to load it
from datasets import load_dataset
ds = load_dataset("katanaml-org/invoices-donut-data-v1")