Improving Text Proposals for Scene Images with Fully Convolutional Networks

TL;DR

This paper enhances text proposal methods for scene images by integrating Fully Convolutional Networks, significantly improving proposal ranking and achieving state-of-the-art results on benchmark datasets.

Contribution

It introduces a novel combination of Text Proposals with Fully Convolutional Networks to improve proposal ranking accuracy in scene text detection.

Findings

Superior performance on ICDAR RRC dataset

Outperforms current state-of-the-art on COCO-text

Improved proposal ranking accuracy

Abstract

Text Proposals have emerged as a class-dependent version of object proposals - efficient approaches to reduce the search space of possible text object locations in an image. Combined with strong word classifiers, text proposals currently yield top state of the art results in end-to-end scene text recognition. In this paper we propose an improvement over the original Text Proposals algorithm of Gomez and Karatzas (2016), combining it with Fully Convolutional Networks to improve the ranking of proposals. Results on the ICDAR RRC and the COCO-text datasets show superior performance over current state-of-the-art.