Donut Invoices

🚀 Fine-tuned Donut Model for Invoice Processing

This project offers a fine-tuned Donut model, specifically optimized for invoice processing, enhancing document understanding and information extraction.

🚀 Quick Start

This repository houses a fine-tuned version of the Donut model, designed for document understanding with a focus on invoice processing. The Donut model, based on the OCR-free Document Understanding Transformer introduced in the paper by Geewok et al., was initially released at https://github.com/clovaai/donut.

The custom fine-tuning aims to boost the Donut model's performance in invoice analysis and extraction. The model was trained on a custom dataset of annotated invoices, which includes several hundred examples. Although the dataset isn't included in this repository, details about its availability will be provided later.

✨ Features

• Targeted Fine-tuning: Optimized specifically for invoice processing to improve accuracy in information extraction.
• Transformer-based Architecture: Leverages self-attention mechanisms for effective document understanding.

📚 Documentation

Model Details

The Donut model uses a transformer-based architecture with self-attention mechanisms for document understanding. By fine-tuning it on a custom invoice dataset, we aim to enhance its ability to accurately extract relevant information from invoices, such as vendor details, billing information, line items, and totals.

Demo

Demo can be found here