TextContourNet: a Flexible and Effective Framework for Improving Scene Text Detection Architecture with a Multi-task Cascade

TL;DR

TextContourNet introduces a multi-task cascade framework that leverages text contour extraction to significantly enhance scene text detection accuracy in natural images.

Contribution

The paper presents a novel CNN-based framework that effectively extracts text contours and integrates them into scene text detection as an auxiliary and cascade task, improving detection performance.

Findings

Contour information improves detection accuracy

Multi-task cascade outperforms single-task methods

Framework achieves state-of-the-art results on benchmarks

Abstract

We study the problem of extracting text instance contour information from images and use it to assist scene text detection. We propose a novel and effective framework for this and experimentally demonstrate that: (1) A CNN that can be effectively used to extract instance-level text contour from natural images. (2) The extracted contour information can be used for better scene text detection. We propose two ways for learning the contour task together with the scene text detection: (1) as an auxiliary task and (2) as multi-task cascade. Extensive experiments with different benchmark datasets demonstrate that both designs improve the performance of a state-of-the-art scene text detector and that a multi-task cascade design achieves the best performance.