donut-invoices

Overview

This repository contains a fine-tuned version of the Donut model for document understanding, specifically tailored for invoice processing. The Donut model is based on the OCR-free Document Understanding Transformer, introduced in the paper by Geewok et al. OCR-free Document Understanding Transformer, and initially released in the repository https://github.com/clovaai/donut.

The purpose of this custom fine-tuning is to enhance the Donut model's performance specifically for invoice analysis and extraction. The model was trained on a custom dataset of annotated invoices, comprising several hundred examples. Although the dataset is not included in this repository, details on its availability will be provided later.

Model Details

The Donut model is a transformer-based architecture that leverages self-attention mechanisms for document understanding. By fine-tuning the model with a custom dataset of invoices, we aim to improve its ability to accurately extract relevant information from invoices, such as vendor details, billing information, line items, and totals.

Demo can be found here