Scene Text Detection for Augmented Reality -- Character Bigram Approach to reduce False Positive Rate

TL;DR

This paper introduces a character bigram approach using a CNN to improve scene text detection in augmented reality, significantly reducing false positives and enhancing detection accuracy efficiently.

Contribution

It presents a novel bigram-based detection method that decreases false positives in scene text spotting with minimal computational overhead.

Findings

Reduces false positive rate by 28.16% on ICDAR 2015 dataset

Detecting bigrams is computationally inexpensive

Improves sliding window text spotters' performance

Abstract

Natural scene text detection is an important aspect of scene understanding and could be a useful tool in building engaging augmented reality applications. In this work, we address the problem of false positives in text spotting. We propose improving the performace of sliding window text spotters by looking for character pairs (bigrams) rather than single characters. An efficient convolutional neural network is designed and trained to detect bigrams. The proposed detector reduces false positive rate by 28.16% on the ICDAR 2015 dataset. We demonstrate that detecting bigrams is a computationally inexpensive way to improve sliding window text spotters.