场景文本检测论文阅读时间线

2019年7月

1. CTPN：文本提议，矩形框，宽度16像素，LSTM，水平矩形文本

Detecting Text in Natural Image withConnectionist Text Proposal Network

2. PSE: 尺度扩展，任意形状，特征金字塔，分割方法

Shape Robust Text Detection with Progressive Scale Expansion Network

3. PMTD: 3D分割掩膜，金字塔掩膜，mask branch

Pyramid Mask Text Detector

4. Mask R-CNN: 实例分割框架，RoIAligen, 28x28 mask

Mask R-CNN

2019年8月

1. CRATD：字符与字符间区域，字符级分割

Character Region Awareness for Text Detection

2. LMTO：直接回归迭代修正，新添网络模块

Look More Than Once: An Accurate Detector for Text of Arbitrary Shapes

2019年9月

1. Corner：角点检测，混合方法，目标检测与实例分割

Multi-Oriented Scene Text Detection via Corner Localization and Region Segmentation

2. Adaptive：自适应文本区域，LSTM，点对

Arbitrary Shape Scene Text Detection with Adaptive Text Region Representation

3. DSSD：SSD加上反卷积模块

Deconvolutional Single Shot Detector

2019年10月

1. Mask TextSpotter：将字符视作分割实例，36个mask

An End-to-End Trainable Neural Network for Spotting Text with Arbitrary Shapes

2. TextCohesion：文本聚集，上下左右像素区域与文本骨架

Detecting Text for Arbitrary Shapes