Chuhui Xue_ECCV2018_Accurate Scene Text Detection through Border Semantics Awareness and Bootstrapping

作者和代码

关键词

文字检测、多方向、FCN、$$xywh\theta$$、multi-stage、border

方法亮点

  • 采用Bootstrapping进行数据扩增
  • 增加border-loss

方法概述

本文方法是直接回归的方法,除了学习text/non-text分类任务,四个点到边界的回归任务(类似EAST),还增加了四条边界的border学习任务,最后输出不是直接用prediction的bounding box,而是用了text score map和四个border map来获得textline。

方法细节

bootstrapping样本扩增

简单说其实就是对文字的polygon做一些重复局部采样,丰富文字patch的多样性。

Fig. 2: Illustration of the bootstrapping based scene text sampling: Given an image with a text line as annotated by the green box, three example text line segments are extracted as highlighted by red boxes where the centers of the sampling windows are taken randomly along the center line of the text line (the shrunk part in yellow color). The rest text regions outside of the sampling windows are filled by inpainting.

具体步骤:

  • 确定中心线,进行shrink 0.1L'
  • 沿中心线随机均匀采样点(确定框中心)
  • 确定框大小(H = 0.9H',W 从[0.2L',2$$d_{min}$$]选取随机值。$$d_{min}$$表示中心点到两条短边的距离的最小值)
  • 把处在原文字框内,但在采样框外(上图的绿色框内红色框外)的部分都进行inpainting(涂抹成单色),以此得到新的训练图和text的groundTruth

是否有进行bootstrapping的效果图如下:

作者认为,进行augmentation之后,feature map上特征更具有一致性,所以distance map更加平滑。

It can be seen that the inclusion of the augmented images helps to produce more consistent text feature maps as well as smoother geometrical distance maps (for regression of text boxes) which leads to more complete instead of broken scene text detections.

我的理解是,一段很长的文本,处在不同段的文字appearance特征可能很不一样。例如Fig 2中有的地方光很强,有的地方光照很弱,因此,bootstrapping采样的是某个文字的segment,越短则这段text的特征越一致,特征学习就会越concentrate在更一致的区域上(其实是把一个学习问题难度降低了)

Fig. 3: The inclusion of augmented images improves the scene text detection: With the inclusion of the augmented images in training, more consistent text feature maps and more complete scene text detections are produced as shown in (d) and (e), as compared with those produced by the baseline model (trained using original training images only) shown in (b) and (c). The coloring in the text feature maps shows the distance information predicted by regressor (blue denotes short distances and red denotes long distance).

增加四个border-pixel的classification

增加4个要学习的border/non-border的classification map。

Fig. 4: Semantics-aware text border detection: Four text border segments are automatically extracted for each text annotation box including a pair of shortedge text border segments in yellow and red colors and a pair of long-edge text border segments in green and blue colors. The four types of text border segments are treated as four types of objects and used to train deep network models, and the trained model is capable of detecting the four types of text border segments as illustrated in Fig. 5c.

具体的四个border的大小如下。上下取0.2H',左右取0.8H',比较特别的是左右的宽度取得比较大,是H'。主要是担心同一文本行文字黏连问题比较多,而且不像上下border具有很长的边(面积大一些)。上下边界主要为了解决多行文字黏连问题。border要学习的是从text到background的transition(...the extracted text border segments capture the transition from text to background or vice versa...)。

作者认为增加短的border之所以可以提升效果是因为处在文字中心部分的像素离两条短边比较远(长条文字),容易产生回归误差,导致检测结果不精确,而增加border像素的loss可以帮助解决这个问题。另一方面,增加两条长border的目的是为了解决挨的比较近的上下两行文本行的黏连问题。

The reason is that text pixels around the middle of texts are far from the text box vertices for long words or text lines which can easily introduce regression errors and lead to inaccurate localization as illustrated in Fig. 5b. At the other end, the long text border segments also help for better scene text detection performance. In particular, the long text border segments can be exploited to separate text lines when neighboring text lines are close to each other.

是否增加border的效果对比如下图:

Fig. 5: The use of semantics-aware text borders improves scene text detection: With the identified text border semantics information as illustrated in (c), scene texts can be localized much more accurately as illustrated in (d) as compared with the detections without using the border semantics information as illustrated in (b). Green boxes give correct detections and red boxes give false detections.

损失函数
  • 总的损失

  • 分类损失:采用DIces Coefficient损失(confidence score of each pixel being a text pixel )

  • 回归损失:采用IOU损失(distances from each pixel to four sides of text boundaries)
  • 边界损失:采用DIces Coefficient损失(confidence score of each pixel being a border pixel)

后处理文本线形成算法
  • 二值化五个map(1个text region map,4个text border map):采用mean_score
  • 算出region map和4个border map的overlap
  • 提取文本行(上下边)
  • 提取左右边界
  • merge四个边界组成的boungding box
  • NMS

实验结果

  • ICDAR2013

  • MSRA-TD500

  • ICDAR2017-RCTW

  • ICDAR2017-MLT

总结与收获

这篇方法的bootstrap来扩增样本的方式很有意思,虽然是在目标检测领域里早有人这么用了,但这是第一次引入到OCR。另外,增加border loss的思路也很直接,与Yue Wu_ICCV2017_Self-Organized Text Detection With Minimal Post-Processing via Border Learning的border有点像。

【论文速读】Chuhui Xue_ECCV2018_Accurate Scene Text Detection through Border Semantics Awareness and Bootstrapping的更多相关文章

  1. 【论文速读】Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation[2018-CPVR]

    方法概述 该方法用一个端到端网络完成文字检测整个过程——除了基础卷积网络(backbone)外,包括两个并行分支和一个后处理.第一个分支是通过一个DSSD网络进行角点检测来提取候选文字区域,第二个分支 ...

  2. 论文速读(Chuhui Xue——【arxiv2019】MSR_Multi-Scale Shape Regression for Scene Text Detection)

    Chuhui Xue--[arxiv2019]MSR_Multi-Scale Shape Regression for Scene Text Detection 论文 Chuhui Xue--[arx ...

  3. 【论文速读】XiangBai_CVPR2018_Rotation-Sensitive Regression for Oriented Scene Text Detection

    XiangBai_CVPR2018_Rotation-Sensitive Regression for Oriented Scene Text Detection 作者和代码 caffe代码 关键词 ...

  4. 论文速读(Yongchao Xu——【2018】TextField_Learning A Deep Direction Field for Irregular Scene Text)

    Yongchao Xu--[2018]TextField_Learning A Deep Direction Field for Irregular Scene Text Detection 论文 Y ...

  5. 论文阅读(Xiang Bai——【arXiv2016】Scene Text Detection via Holistic, Multi-Channel Prediction)

    Xiang Bai--[arXiv2016]Scene Text Detection via Holistic, Multi-Channel Prediction 目录 作者和相关链接 方法概括 创新 ...

  6. 论文阅读(Weilin Huang——【TIP2016】Text-Attentional Convolutional Neural Network for Scene Text Detection)

    Weilin Huang--[TIP2015]Text-Attentional Convolutional Neural Network for Scene Text Detection) 目录 作者 ...

  7. XiangBai——【CVPR2018】Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation

    XiangBai——[CVPR2018]Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentat ...

  8. Learning Markov Clustering Networks for Scene Text Detection

    Learning Markov Clustering Networks for Scene Text Detection 论文下载:https://arxiv.org/pdf/1805.08365v1 ...

  9. 【论文速读】ChengLin_Liu_ICCV2017_Deep_Direct_Regression_for_Multi-Oriented_Scene_Text_Detection

    ChengLin Liu_ICCV2017_Deep Direct Regression for Multi-Oriented Scene Text Detection 作者 关键词 文字检测.多方向 ...

随机推荐

  1. 执行Python出现LookupError: unknown encoding: cp65001解决办法

    执行Python出现LookupError: unknown encoding: cp65001错误 dos下执行以下命令即可 chcp 以上.

  2. dagger2 依赖注入

    前言: 当 mvp + dagger2 + rxjava 三者遇上,架构更加清晰,代码更加灵活,巧妙结合. 依赖注入:是为了解耦,达到高内聚低耦合的目的,保证代码的健壮性.灵活性和可维护性. publ ...

  3. 【转】Spring总结以及在面试中的一些问题

    [转]Spring总结以及在面试中的一些问题. 1.谈谈你对spring IOC和DI的理解,它们有什么区别? IoC Inverse of Control 反转控制的概念,就是将原本在程序中手动创建 ...

  4. 编写程序,将来自文件中的行保存在一个vector<string>,然后使用一个istringstream 从vector中读取数据,每次读一个单词

    #include<fstream> #include <vector> #include<string> #include<iostream> #inc ...

  5. vue_class 绑定_style 绑定

    1. class 绑定 data: { myClass: "bClass", hasA: true, hasB: false } 字符串模式: 类名不确定 <p class= ...

  6. 什么是Hash?Hash有哪些特性?

    Hash 把任意长度的输入通过散列算法变换成固定长度的输出 Hash的特性: 输入域无穷,输出域有限.例如:有无穷多个(在工程中可以具体到多少个,例如1000)输入参数经过hash函数映射后得到有限的 ...

  7. hadoop本地开发环境搭建

    1:下载hadoop2.7.3并解压 2:配置hadoop2.7.3环境变量 HADOOP_HOME %HADOOP_HOME%\bin 3:下载hadoop-eclipse-plugin插件 网址: ...

  8. java 实现文件上传下载以及查看

    项目的目录结构 代码  IOUtils.java package cn.edu.zyt.util; import java.io.IOException; import java.io.InputSt ...

  9. WinForm打包或部署

    一.新建InstallShield项目 二. 三. 四. 五. 六. 七. 最后重新生成,安装包一般在E\Setup1\Setup1\Express\SingleImage\DiskImages\DI ...

  10. 数据仓库 - 2.数据仓库设计思路及ETL设计思路

    一.数据仓库构建思想 构造数据仓库有两种方式:一是自上而下,一是自下而上. Bill Inmon先生推崇“自上而下”的方式,即一个企业建立唯一的数据中心,就像一个数据的仓库,其中数据是经过整合.经过清 ...