ExTTNet: A Deep Learning Algorithm for Extracting Table Texts from Invoice Images

TL;DR

This paper introduces ExTTNet, a deep learning-based method that extracts product table texts from invoice images by combining OCR with feature extraction and neural network classification, achieving high accuracy.

Contribution

The study presents a novel deep learning approach, ExTTNet, that improves table text extraction accuracy from invoice images by integrating OCR and feature-based neural network classification.

Findings

F1 score of 0.92 achieved

Effective combination of OCR and neural network

Training completed in 162 minutes

Abstract

In this work, product tables in invoices are obtained autonomously via a deep learning model, which is named as ExTTNet. Firstly, text is obtained from invoice images using Optical Character Recognition (OCR) techniques. Tesseract OCR engine [37] is used for this process. Afterwards, the number of existing features is increased by using feature extraction methods to increase the accuracy. Labeling process is done according to whether each text obtained as a result of OCR is a table element or not. In this study, a multilayer artificial neural network model is used. The training has been carried out with an Nvidia RTX 3090 graphics card and taken $162$ minutes. As a result of the training, the F1 score is $0.92$.