Segmentation-Free Approaches for Handwritten Numeral String Recognition

TL;DR

This paper introduces segmentation-free deep learning methods for recognizing handwritten numeral strings of unknown length, demonstrating state-of-the-art performance without over-segmentation issues and highlighting the importance of contextual information.

Contribution

The paper proposes novel segmentation-free approaches using CNNs and contextual length classifiers for handwritten numeral string recognition.

Findings

Segmentation-free methods achieve state-of-the-art accuracy

Synthetic dataset effectively trains end-to-end models

Contextual length classification improves recognition performance

Abstract

This paper presents segmentation-free strategies for the recognition of handwritten numeral strings of unknown length. A synthetic dataset of touching numeral strings of sizes 2-, 3- and 4-digits was created to train end-to-end solutions based on Convolutional Neural Networks. A robust experimental protocol is used to show that the proposed segmentation-free methods may reach the state-of-the-art performance without suffering the heavy burden of over-segmentation based methods. In addition, they confirmed the importance of introducing contextual information in the design of end-to-end solutions, such as the proposed length classifier when recognizing numeral strings.