An Image Dataset of Text Patches in Everyday Scenes

TL;DR

This paper introduces COCO-Text-Patch, a large dataset of small images of text in everyday scenes, designed to improve text detection and recognition systems in natural images, with initial deep learning benchmarks showing high accuracy.

Contribution

The paper presents a new dataset of 354,000 small images for text verification in natural scenes and demonstrates its utility by training deep networks achieving over 90% accuracy.

Findings

Deep networks trained on the dataset achieved over 90% accuracy.

The dataset supports development of automated text detection in social media images.

All resources including images, code, and models are publicly available.

Abstract

This paper describes a dataset containing small images of text from everyday scenes. The purpose of the dataset is to support the development of new automated systems that can detect and analyze text. Although much research has been devoted to text detection and recognition in scanned documents, relatively little attention has been given to text detection in other types of images, such as photographs that are posted on social-media sites. This new dataset, known as COCO-Text-Patch, contains approximately 354,000 small images that are each labeled as "text" or "non-text". This dataset particularly addresses the problem of text verification, which is an essential stage in the end-to-end text detection and recognition pipeline. In order to evaluate the utility of this dataset, it has been used to train two deep convolution neural networks to distinguish text from non-text. One network is inspired by the GoogLeNet architecture, and the second one is based on CaffeNet. Accuracy levels of 90.2% and 90.9% were obtained using the two networks, respectively. All of the images, source code, and deep-learning trained models described in this paper will be publicly available