知识荟萃

OCR文字，车牌，验证码识别专知荟萃

入门学习

端到端的OCR：基于CNN的实现
- blog: [http://blog.xlvector.net/2016-05/mxnet-ocr-cnn/]
如何用卷积神经网络CNN识别手写数字集？
- blog: [http://www.cnblogs.com/charlotte77/p/5671136.html]
OCR文字识别用的是什么算法？
- [https://www.zhihu.com/question/20191727]
基于计算机视觉/深度学习打造先进OCR工作流 Creating a Modern OCR Pipeline Using Computer Vision and Deep Learning
- [https://blogs.dropbox.com/tech/2017/04/creating-a-modern-ocr-pipeline-using-computer-vision-and-deep-learning/]
车牌识别中的不分割字符的端到端(End-to-End)识别
- [http://m.blog.csdn.net/Relocy/article/details/52174198]
端到端的OCR：基于CNN的实现
- [http://blog.xlvector.net/2016-05/mxnet-ocr-cnn/]
腾讯OCR—自动识别技术，探寻文字真实的容颜
- [http://blog.xlvector.net/2016-05/mxnet-ocr-cnn/]
Tesseract-OCR引擎入门
- [http://blog.csdn.net/xiaochunyong/article/details/7193744]
汽车挡风玻璃VIN码识别
- [https://github.com/DoctorDYL/VINOCR]
车牌识别算法的关键技术及其研究现状
- [http://www.siat.cas.cn/xscbw/xsqk/201012/W020101222564768411838.pdf]
端到端的OCR：验证码识别
- [https://zhuanlan.zhihu.com/p/21344595?refer=xlvector]

论文及代码

文字识别

Multi-digit Number Recognition from Street View Imagery using Deep Convolutional Neural Networks
- intro: Google. Ian J. Goodfellow
- arxiv: [https://arxiv.org/abs/1312.6082]
End-to-End Text Recognition with Convolutional Neural Networks
- paper: [http://www.cs.stanford.edu/~acoates/papers/wangwucoatesng_icpr2012.pdf]
- PhD thesis: [http://cs.stanford.edu/people/dwu4/HonorThesis.pdf]
Word Spotting and Recognition with Embedded Attributes
- paper: [http://ieeexplore.ieee.org.sci-hub.org/xpl/articleDetails.jsp?arnumber=6857995&filter%3DAND%28p_IS_Number%3A6940341%29]
Reading Text in the Wild with Convolutional Neural Networks
- arxiv: [http://arxiv.org/abs/1412.1842]
- homepage: [http://www.robots.ox.ac.uk/~vgg/publications/2016/Jaderberg16/]
- demo: [http://zeus.robots.ox.ac.uk/textsearch/#/search/]
- code: [http://www.robots.ox.ac.uk/~vgg/research/text/]
Deep structured output learning for unconstrained text recognition
- arxiv: [http://arxiv.org/abs/1412.5903]
Deep Features for Text Spotting
- paper: [http://www.robots.ox.ac.uk/~vgg/publications/2014/Jaderberg14/jaderberg14.pdf]
- bitbucket: [https://bitbucket.org/jaderberg/eccv2014_textspotting]
- gitxiv: [http://gitxiv.com/posts/uB4y7QdD5XquEJ69c/deep-features-for-text-spotting]
Reading Scene Text in Deep Convolutional Sequences
- arxiv: [http://arxiv.org/abs/1506.04395]
DeepFont: Identify Your Font from An Image
- arxiv: [http://arxiv.org/abs/1507.03196]
An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition
- intro: Convolutional Recurrent Neural Network
- arxiv: [http://arxiv.org/abs/1507.05717]
- github: [https://github.com/bgshih/crnn]
- github: [https://github.com/meijieru/crnn.pytorch]
Recursive Recurrent Nets with Attention Modeling for OCR in the Wild
- arxiv: [http://arxiv.org/abs/1603.03101]
Writer-independent Feature Learning for Offline Signature Verification using Deep Convolutional Neural Networks
- arxiv: [http://arxiv.org/abs/1604.00974]
DeepText: A Unified Framework for Text Proposal Generation and Text Detection in Natural Images
- arxiv: [http://arxiv.org/abs/1605.07314]
End-to-End Interpretation of the French Street Name Signs Dataset
- paper: [http://link.springer.com/chapter/10.1007%2F978-3-319-46604-0_30]
- github: [https://github.com/tensorflow/models/tree/master/street]
End-to-End Subtitle Detection and Recognition for Videos in East Asian Languages via CNN Ensemble with Near-Human-Level Performance
- arxiv: [https://arxiv.org/abs/1611.06159]
Smart Library: Identifying Books in a Library using Richly Supervised Deep Scene Text Reading
- arxiv: [https://arxiv.org/abs/1611.07385]
Improving Text Proposals for Scene Images with Fully Convolutional Networks
- intro: Universitat Autonoma de Barcelona & University of Florence
- intro: International Conference on Pattern Recognition - DLPR workshop
- arxiv: [https://arxiv.org/abs/1702.05089]
Scene Text Eraser
- [https://arxiv.org/abs/1705.02772]
Attention-based Extraction of Structured Information from Street View Imagery
- intro: University College London & Google Inc
- arxiv: [https://arxiv.org/abs/1704.03549]
- github: [https://github.com/tensorflow/models/tree/master/attention_ocr]
STN-OCR: A single Neural Network for Text Detection and Text Recognition
- arxiv: [https://arxiv.org/abs/1707.08831]
- github: [https://github.com/Bartzi/stn-ocr]
Sequence to sequence learning for unconstrained scene text recognition
- intro: master thesis
- arxiv: [http://arxiv.org/abs/1607.06125]
Drawing and Recognizing Chinese Characters with Recurrent Neural Network
- arxiv: [https://arxiv.org/abs/1606.06539]
Learning Spatial-Semantic Context with Fully Convolutional Recurrent Network for Online Handwritten Chinese Text Recognition
- intro: correct rates: Dataset-CASIA 97.10% and Dataset-ICDAR 97.15%
- arxiv: [https://arxiv.org/abs/1610.02616]
Stroke Sequence-Dependent Deep Convolutional Neural Network for Online Handwritten Chinese Character Recognition
- arxiv: [https://arxiv.org/abs/1610.04057]
Visual attention models for scene text recognition
- [https://arxiv.org/abs/1706.01487]
Focusing Attention: Towards Accurate Text Recognition in Natural Images
- intro: ICCV 2017
- arxiv: [https://arxiv.org/abs/1709.02054]
Scene Text Recognition with Sliding Convolutional Character Models
- [https://arxiv.org/abs/1709.01727]
AdaDNNs: Adaptive Ensemble of Deep Neural Networks for Scene Text Recognition
- [https://arxiv.org/abs/1710.03425]
A New Hybrid-parameter Recurrent Neural Networks for Online Handwritten Chinese Character Recognition
- [https://arxiv.org/abs/1711.02809]
Arbitrarily-Oriented Text Recognition
- intro: A method used in ICDAR 2017 word recognition competitions
- arxiv: [https://arxiv.org/abs/1711.04226]

文字检测

Object Proposals for Text Extraction in the Wild
- intro: ICDAR 2015
- arxiv: [http://arxiv.org/abs/1509.02317]
- github: [https://github.com/lluisgomez/TextProposals]
Text-Attentional Convolutional Neural Networks for Scene Text Detection
- arxiv: [http://arxiv.org/abs/1510.03283]
Accurate Text Localization in Natural Image with Cascaded Convolutional Text Network
- arxiv: [http://arxiv.org/abs/1603.09423]
Synthetic Data for Text Localisation in Natural Images
- intro: CVPR 2016
- project page: [http://www.robots.ox.ac.uk/~vgg/data/scenetext/]
- arxiv: [http://arxiv.org/abs/1604.06646]
- paper: [http://www.robots.ox.ac.uk/~vgg/data/scenetext/gupta16.pdf]
- github: [https://github.com/ankush-me/SynthText]
Scene Text Detection via Holistic, Multi-Channel Prediction
- arxiv: [http://arxiv.org/abs/1606.09002]
Detecting Text in Natural Image with Connectionist Text Proposal Network
- intro: ECCV 2016
- arxiv: [http://arxiv.org/abs/1609.03605]
- github: [https://github.com/tianzhi0549/CTPN]
- github: [https://github.com/qingswu/CTPN]
- demo: [http://textdet.com/]
- github: [https://github.com/eragonruan/text-detection-ctpn]
TextBoxes: A Fast Text Detector with a Single Deep Neural Network
- intro: AAAI 2017
- arxiv: [https://arxiv.org/abs/1611.06779]
- github: [https://github.com/MhLiao/TextBoxes]
- github: [https://github.com/xiaodiu2010/TextBoxes-TensorFlow]
Deep Matching Prior Network: Toward Tighter Multi-oriented Text Detection
- intro: CVPR 2017
- intro: F-measure 70.64%, outperforming the existing state-of-the-art method with F-measure 63.76%
- arxiv: [https://arxiv.org/abs/1703.01425]
Detecting Oriented Text in Natural Images by Linking Segments
- intro: CVPR 2017
- arxiv: [https://arxiv.org/abs/1703.06520]
- github: [https://github.com/dengdan/seglink]
Deep Direct Regression for Multi-Oriented Scene Text Detection
- arxiv: [https://arxiv.org/abs/1703.08289]
Cascaded Segmentation-Detection Networks for Word-Level Text Spotting
- [https://arxiv.org/abs/1704.00834]
WordFence: Text Detection in Natural Images with Border Awareness
- intro: ICIP 2017
- arcxiv: [https://arxiv.org/abs/1705.05483]
SSD-text detection: Text Detector
- intro: A modified SSD model for text detection
- github: [https://github.com/oyxhust/ssd-text_detection]
R2CNN: Rotational Region CNN for Orientation Robust Scene Text Detection
- intro: Samsung R&D Institute China
- arxiv: [https://arxiv.org/abs/1706.09579]
R-PHOC: Segmentation-Free Word Spotting using CNN
- intro: ICDAR 2017
- arxiv: [https://arxiv.org/abs/1707.01294]
Towards End-to-end Text Spotting with Convolutional Recurrent Neural Networks
- [https://arxiv.org/abs/1707.03985]
EAST: An Efficient and Accurate Scene Text Detector
- intro: CVPR 2017
- arxiv: [https://arxiv.org/abs/1704.03155]
- github: [https://github.com/argman/EAST]
Deep Scene Text Detection with Connected Component Proposals
- intro: Amap Vision Lab, Alibaba Group
- arxiv: [https://arxiv.org/abs/1708.05133]
Single Shot Text Detector with Regional Attention
- intro: ICCV 2017
- arxiv: [https://arxiv.org/abs/1709.00138]
- github: [https://github.com/BestSonny/SSTD]
- code: [http://sstd.whuang.org]
Fused Text Segmentation Networks for Multi-oriented Scene Text Detection
- [https://arxiv.org/abs/1709.03272]
Deep Residual Text Detection Network for Scene Text
- intro: IAPR International Conference on Document Analysis and Recognition 2017. Samsung R&D Institute of China, Beijing
- arxiv: [https://arxiv.org/abs/1711.04147]
Feature Enhancement Network: A Refined Scene Text Detector
- intro: AAAI 2018
- arxiv: [https://arxiv.org/abs/1711.04249]
ArbiText: Arbitrary-Oriented Text Detection in Unconstrained Scene
- [https://arxiv.org/abs/1711.11249]

验证码破解

Using deep learning to break a Captcha system
- intro: "Using Torch code to break simplecaptcha with 92% accuracy"
- blog: [https://deepmlblog.wordpress.com/2016/01/03/how-to-break-a-captcha-system/]
- github: [https://github.com/arunpatala/captcha]
Breaking reddit captcha with 96% accuracy
- blog: [https://deepmlblog.wordpress.com/2016/01/05/breaking-reddit-captcha-with-96-accuracy/]
- github: [https://github.com/arunpatala/reddit.captcha]
I’m not a human: Breaking the Google reCAPTCHA
- paper: [https://www.blackhat.com/docs/asia-16/materials/asia-16-Sivakorn-Im-Not-a-Human-Breaking-the-Google-reCAPTCHA-wp.pdf]
Neural Net CAPTCHA Cracker
- slides: [http://www.cs.sjsu.edu/faculty/pollett/masters/Semesters/Spring15/geetika/CS298%20Slides%20-%20PDF]
- github: [https://github.com/bgeetika/Captcha-Decoder]
- demo: [http://cp-training.appspot.com/]
Recurrent neural networks for decoding CAPTCHAS
Reading irctc captchas with 95% accuracy using deep learning
- github: [https://github.com/arunpatala/captcha.irctc]
I Am Robot: Learning to Break Semantic Image CAPTCHAs
- intro: automatically solving 70.78% of the image reCaptchachallenges, while requiring only 19 seconds per challenge. apply to the Facebook image captcha and achieve an accuracy of 83.5%
- paper: [http://www.cs.columbia.edu/~polakis/papers/sivakorn_eurosp16.pdf]
SimGAN-Captcha
- intro: Solve captcha without manually labeling a training set
- github: [https://github.com/rickyhan/SimGAN-Captcha]

手写体识别

High Performance Offline Handwritten Chinese Character Recognition Using GoogLeNet and Directional Feature Maps
- arxiv: [http://arxiv.org/abs/1505.04925]
- github: [https://github.com/zhongzhuoyao/HCCR-GoogLeNet]
Recognize your handwritten numbers
- [https://medium.com/@o.kroeger/recognize-your-handwritten-numbers-3f007cbe46ff#.jllz62xgu]
Handwritten Digit Recognition using Convolutional Neural Networks in Python with Keras
- blog: [http://machinelearningmastery.com/handwritten-digit-recognition-using-convolutional-neural-networks-python-keras/]
MNIST Handwritten Digit Classifier
- github: [https://github.com/karandesai-96/digit-classifier]
LeNet – Convolutional Neural Network in Python
- blog: [http://www.pyimagesearch.com/2016/08/01/lenet-convolutional-neural-network-in-python/]
Scan, Attend and Read: End-to-End Handwritten Paragraph Recognition with MDLSTM Attention
- arxiv: [http://arxiv.org/abs/1604.03286]
MLPaint: the Real-Time Handwritten Digit Recognizer
Training a Computer to Recognize Your Handwriting
- [https://medium.com/@annalyzin/training-a-computer-to-recognize-your-handwriting-24b808fb584#.gd4pb9jk2]
Using TensorFlow to create your own handwriting recognition engine
- blog: [https://niektemme.com/2016/02/21/tensorflow-handwriting/]
- github: [https://github.com/niektemme/tensorflow-mnist-predict/]
Building a Deep Handwritten Digits Classifier using Microsoft Cognitive Toolkit
- blog: [https://medium.com/@tuzzer/building-a-deep-handwritten-digits-classifier-using-microsoft-cognitive-toolkit-6ae966caec69#.c3h6o7oxf]
- github: [https://github.com/tuzzer/ai-gym/blob/a97936619cf56b5ed43329c6fa13f7e26b1d46b8/MNIST/minist_softmax_cntk.py]
Hand Writing Recognition Using Convolutional Neural Networks
- intro: This CNN-based model for recognition of hand written digits attains a validation accuracy of 99.2% after training for 12 epochs. Its trained on the MNIST dataset on Kaggle.
- github: [https://github.com/ayushoriginal/HandWritingRecognition-CNN]
Design of a Very Compact CNN Classifier for Online Handwritten Chinese Character Recognition Using DropWeight and Global Pooling
- intro: 0.57 MB, performance is decreased only by 0.91%.
- arxiv: [https://arxiv.org/abs/1705.05207]
Handwritten digit string recognition by combination of residual network and RNN-CTC
- [https://arxiv.org/abs/1710.03112]

车牌识别

Reading Car License Plates Using Deep Convolutional Neural Networks and LSTMs
- arxiv: [http://arxiv.org/abs/1601.05610]
Number plate recognition with Tensorflow
- blog: [http://matthewearl.github.io/2016/05/06/cnn-anpr/]
- github: [https://github.com/matthewearl/deep-anpr]
end-to-end-for-plate-recognition
- github: [https://github.com/szad670401/end-to-end-for-chinese-plate-recognition]
Segmentation-free Vehicle License Plate Recognition using ConvNet-RNN
- intro: International Workshop on Advanced Image Technology, January, 8-10, 2017. Penang, Malaysia. Proceeding IWAIT2017
- arxiv: [https://arxiv.org/abs/1701.06439]
License Plate Detection and Recognition Using Deeply Learned Convolutional Neural Networks
- arxiv: [https://arxiv.org/abs/1703.07330]
- api: [https://www.sighthound.com/products/cloud]
Adversarial Generation of Training Examples for Vehicle License Plate Recognition
- [https://arxiv.org/abs/1707.03124]
Towards End-to-End Car License Plates Detection and Recognition with Deep Neural Networks
- [https://arxiv.org/abs/1709.08828]

实战项目

多标签分类,端到端基于mxnet的中文车牌识别
- [https://github.com/szad670401/end-to-end-for-chinese-plate-recognition]
中国二代身份证光学识别
- [https://github.com/KevinGong2013/ChineseIDCardOCR]
EasyPR 一个开源的中文车牌识别系统
- [https://github.com/liuruoze/EasyPR]
汽车挡风玻璃VIN码识别
- [https://github.com/DoctorDYL/VINOCR]
CLSTM : A small C++ implementation of LSTM networks, focused on OCR
- github: [https://github.com/tmbdev/clstm]
OCR text recognition using tensorflow with attention
- github: [https://github.com/pannous/caffe-ocr]
- github: [https://github.com/pannous/tensorflow-ocr]
Digit Recognition via CNN: digital meter numbers detection
- github: [https://github.com/SHUCV/digit]
Attention-OCR: Visual Attention based OCR
- github: [https://github.com/da03/Attention-OCR]
umaru: An OCR-system based on torch using the technique of LSTM/GRU-RNN, CTC and referred to the works of rnnlib and clstm
- github: [https://github.com/edward-zhu/umaru]
Tesseract.js: Pure Javascript OCR for 62 Languages
- homepage: [http://tesseract.projectnaptha.com/]
- github: [https://github.com/naptha/tesseract.js]
DeepHCCR: Offline Handwritten Chinese Character Recognition based on GoogLeNet and AlexNet
- github: [https://github.com/chongyangtao/DeepHCCR]
deep ocr: make a better chinese character recognition OCR than tesseract
- [https://github.com/JinpengLI/deep_ocr]
Practical Deep OCR for scene text using CTPN + CRNN
- [https://github.com/AKSHAYUBHAT/DeepVideoAnalytics/blob/master/notebooks/OCR/readme.md]
Text-Detection-using-py-faster-rcnn-framework
- github: [https://github.com/jugg1024/Text-Detection-with-FRCN]
ocropy: Python-based tools for document analysis and OCR
- github: [https://github.com/tmbdev/ocropy]
Extracting text from an image using Ocropus
- blog: [http://www.danvk.org/2015/01/09/extracting-text-from-an-image-using-ocropus.html]

视频

LSTMs for OCR

youtube: [https://www.youtube.com/watch?v=5vW8faXvnrc]

初步版本，水平有限，有错误或者不完善的地方，欢迎大家提建议和补充，会一直保持更新，本文为专知内容组原创内容，未经允许不得转载，如需转载请发送邮件至fangquanyi@gmail.com 或联系微信专知小助手（Rancho_Fang）

敬请关注http://www.zhuanzhi.ai 和关注专知公众号，获取第一手AI相关知识

VIP内容

ICDAR · DUTTA Anjan · 模式识别 · 计算机视觉 · OCR（光学字符识别） ·

2019 年 9 月 20 日

【ICDAR2019教程】模式识别和文档图像中基于图的方法，Graph-based Methods in Pattern Recognition and Document Image Analysis

专知会员服务

专知，提供专业可信的知识分发服务，让认知协作更快更好！

主题： Graph-based Methods in Pattern Recognition and Document Image Analysis

简介： 模式识别和文档图像分析中的许多任务被公式化为图形匹配问题。尽管问题具有NP难性，但快速准确的收敛已在模式识别的广泛应用中取得了重大进展。因此，学习基于图的表示形式和相关技术是真正兴趣。在本教程中，我们将介绍用于获得不同应用程序的图形表示的许多方法。之后，我们将解释用于在图域中识别，分类，检测和许多其他任务的基于图的不同算法，方法和技术。我们将介绍最近的趋势，包括图卷积网络和图中的消息传递，重点介绍在各种模式识别问题中的应用，例如化学分子分类和网络图形表示中的检测。此外，除了这些算法在文档图像分析和识别（尤其是模式识别）领域的不同应用之外，还将提供相关经验。

嘉宾介绍： DUTTA Anjan是位于巴塞罗那计算机视觉中心的P-SPHERE项目下的Marie-Curie博士后。他于2014年获得巴塞罗那自治大学（UAB）的计算机科学博士学位。他是IJCV，IEEE TCYB，IEEE TNNLS，PR，PRL等期刊的定期审稿人，并经常担任BMVC，ICPR，ACPR和ICFHR等各种科学会议的程序委员会委员。他最近的研究兴趣围绕视觉对象的基于图形的表示和解决计算机视觉，模式识别和机器学习中各种任务的基于图形的算法。

Luqman Muhammad Muzzamil博士是文档图像分析，模式识别和计算机视觉的研究科学家。自2015年11月以来，卢克曼目前在拉罗谢尔大学（法国）的L3i实验室担任研究工程师。Luqman曾在波尔多生物信息学中心（波尔多生物信息中心）担任研究工程师，并在拉罗谢尔大学（法国）的L3i实验室担任Jean-Marc Ogier教授的博士后研究员。 Luqman拥有FrançoisRabelais的图尔大学（法国）和巴塞罗那的Autonoma大学（西班牙）的计算机科学博士学位。他的博士学位论文由Jean-Yves Ramel教授和Josep Llados教授共同指导。他的研究兴趣包括结构模式识别，文档图像分析，基于相机的文档分析和识别，图形识别，机器学习，计算机视觉，增强现实和仿生学。

成为VIP会员查看完整内容