DeepText: A Unified Framework for Text Proposal Generation and Text Detection in Natural Images

TL;DR

DeepText introduces a unified CNN-based framework for text proposal generation and detection in natural images, significantly improving recall and accuracy over previous methods.

Contribution

The paper presents a novel unified CNN framework with inception-RPN and multilevel pooling for efficient text detection and high recall in natural images.

Findings

Achieves F-measure of 0.83 on ICDAR 2011

Achieves F-measure of 0.85 on ICDAR 2013

Outperforms previous state-of-the-art methods

Abstract

In this paper, we develop a novel unified framework called DeepText for text region proposal generation and text detection in natural images via a fully convolutional neural network (CNN). First, we propose the inception region proposal network (Inception-RPN) and design a set of text characteristic prior bounding boxes to achieve high word recall with only hundred level candidate proposals. Next, we present a powerful textdetection network that embeds ambiguous text category (ATC) information and multilevel region-of-interest pooling (MLRP) for text and non-text classification and accurate localization. Finally, we apply an iterative bounding box voting scheme to pursue high recall in a complementary manner and introduce a filtering algorithm to retain the most suitable bounding box, while removing redundant inner and outer boxes for each text instance. Our approach achieves an F-measure of 0.83 and 0.85 on the ICDAR 2011 and 2013 robust text detection benchmarks, outperforming previous state-of-the-art results.