基于多边形偏移蒙版和边界增强的场景文本检测

Scene text detection based polygon offset mask and border augmentation

目前,多方向文本检测方法已经在各种数据集上取得了不错的性能,但是任意形状文本检测仍然存在一些困难,尤其是具有不同大小、形状、方向、颜色和样式的文本实例。为了更好地区分连续任意形状的文本实例和周边非文本区域,提出了一种基于分段的文本检测器,通过使用多边形偏移蒙版和边界增强来检测任意形状的场景文本。为了评估该方法的有效性,在ICDAR2015和Total-Text等公开数据集上进行了多组对比实验,实验结果证明该方法有着更卓越的性能。

At present,multi-oriented text detection methods have achieved good performance on various benchmarks,but there are still some difficulties in detecting arbitrary shaped text,especially text instances with different sizes,shapes,directions,colors,and styles.In order to better distinguish between continuous arbitrary-shaped text instances and surrounding non-text regions,this paper proposed a segmentation-based text detector that used polygon offset mask and border augmentation to detect arbitrary shaped scene text.In order to evaluate the effectiveness of the proposed method,this paper conducted multiple sets of comparative experiments on public benchmarks such as ICDAR2015 and Total-Text dataset.The experimental results prove that the proposed method has more excellent performance.