Data Efficient Training of a U-Net Based Architecture for Structured Documents Localization

TL;DR

This paper introduces SDL-Net, a U-Net based architecture for structured document localization that is pre-trainable on generic datasets, enabling efficient fine-tuning for new document classes with limited labeled data.

Contribution

The paper presents SDL-Net, a novel, data-efficient U-Net like architecture that improves document localization by leveraging pre-training and fast fine-tuning on new document classes.

Findings

SDL-Net achieves high localization accuracy on proprietary datasets.

Pre-training on generic datasets enhances generalization to new document types.

Fine-tuning requires less data and computational resources.

Abstract

Structured documents analysis and recognition are essential for modern online on-boarding processes, and document localization is a crucial step to achieve reliable key information extraction. While deep-learning has become the standard technique used to solve document analysis problems, real-world applications in industry still face the limited availability of labelled data and of computational resources when training or fine-tuning deep-learning models. To tackle these challenges, we propose SDL-Net: a novel U-Net like encoder-decoder architecture for the localization of structured documents. Our approach allows pre-training the encoder of SDL-Net on a generic dataset containing samples of various document classes, and enables fast and data-efficient fine-tuning of decoders to support the localization of new document classes. We conduct extensive experiments on a proprietary dataset of structured document images to demonstrate the effectiveness and the generalization capabilities of the proposed approach.