Modular Multimodal Architecture for Document Classification

TL;DR

This paper introduces a modular multimodal architecture that combines visual and textual features to improve document page classification, achieving state-of-the-art accuracy on the RVL-CDIP benchmark.

Contribution

The paper presents a novel multimodal approach that integrates visual and textual information for document classification, surpassing existing methods.

Findings

Achieved 93.03% accuracy on RVL-CDIP benchmark.

Outperformed previous state-of-the-art methods.

Demonstrated effectiveness of multimodal integration in document analysis.

Abstract

Page classification is a crucial component to any document analysis system, allowing for complex branching control flows for different components of a given document. Utilizing both the visual and textual content of a page, the proposed method exceeds the current state-of-the-art performance on the RVL-CDIP benchmark at 93.03% test accuracy.