Curved Text Detection in Natural Scene Images with Semi- and Weakly-Supervised Learning

TL;DR

This paper introduces a semi- and weakly-supervised learning approach for curved text detection in natural scene images, significantly reducing the need for pixel-level annotations while maintaining high accuracy.

Contribution

It presents a novel training scheme that leverages limited pixel-level annotations and abundant weakly labeled data, including unlabeled data, to train effective text detectors.

Findings

Achieves comparable performance to state-of-the-art with only 10% pixel annotations.

Effectively utilizes bounding boxes to generate pseudo masks for weak supervision.

Demonstrates strong generalization across different datasets.

Abstract

Detecting curved text in the wild is very challenging. Recently, most state-of-the-art methods are segmentation based and require pixel-level annotations. We propose a novel scheme to train an accurate text detector using only a small amount of pixel-level annotated data and a large amount of data annotated with rectangles or even unlabeled data. A baseline model is first obtained by training with the pixel-level annotated data and then used to annotate unlabeled or weakly labeled data. A novel strategy which utilizes ground-truth bounding boxes to generate pseudo mask annotations is proposed in weakly-supervised learning. Experimental results on CTW1500 and Total-Text demonstrate that our method can substantially reduce the requirement of pixel-level annotated data. Our method can also generalize well across two datasets. The performance of the proposed method is comparable with the state-of-the-art methods with only 10% pixel-level annotated data and 90% rectangle-level weakly annotated data.