GithubHelp home page GithubHelp logo

Comments (17)

wanghaisheng avatar wanghaisheng commented on June 10, 2024 7

@rolai 自己造吧
印刷体的话

行数据的话 可以参考
https://github.com/wanghaisheng/awesome-ocr/wiki/Training-an-Ocropus-OCR-model-中文
单字的数据可以参考
https://github.com/szad670401/OCR_CharGen
https://github.com/szad670401/CNN_Chinese_ocr_training

手写的话 可以参考
https://github.com/tianrolin/HCCR-ResNet

from awesome-ocr.

wanghaisheng avatar wanghaisheng commented on June 10, 2024

https://github.com/shekkizh/FCN.tensorflow
https://github.com/yulongpo/label-picture
https://github.com/MarvinTeichmann/tensorflow-fcn
https://github.com/TrafficObjectDetection/Parallel-RFCN Faster R-CNN in MXNet with distributed implementation and data parallelization

from awesome-ocr.

rolai avatar rolai commented on June 10, 2024

正要做这个领域,谢谢楼主分享~
请问楼主有没有开源的标注数据可以下载呢?

from awesome-ocr.

hsddlz avatar hsddlz commented on June 10, 2024

感谢楼主,请问我现在想通过fasterRcnn训练手机图片OCR ,我的标注都是一系列词语,对于OCR识别,我应该是否直接采用fasterRCNN来分几万类??

from awesome-ocr.

wanghaisheng avatar wanghaisheng commented on June 10, 2024

@hsddlz fastRCNN 我没用过 给不出你答案
你可以尝试问问这个#54

from awesome-ocr.

willard-yuan avatar willard-yuan commented on June 10, 2024

@hsddlz 不要那样做,是用faster-RCNN来做文字定位,然后再采用LSTM或者CTC来做检测的区域做识别部分。

from awesome-ocr.

xiaomaxiao avatar xiaomaxiao commented on June 10, 2024

CTPN 对汉字定位还不错, 就是没有放train细节

from awesome-ocr.

wanghaisheng avatar wanghaisheng commented on June 10, 2024

@xiaomaxiao 求url

from awesome-ocr.

xiaomaxiao avatar xiaomaxiao commented on June 10, 2024

https://github.com/tianzhi0549/CTPN

这个,用训练好的模型,汉字定位还挺准的, 就是没有放train的部分。

from awesome-ocr.

violet-day avatar violet-day commented on June 10, 2024

@wanghaisheng hello,请教下楼主,如果是针对营业执照这样的场景,使用faster-rcnn识别区域,bounding box有什么原则嘛,是基于word,还是基于line,是不是图片中所有文字区域都要标记出来(因为比如营业范围可能会有很多行,比较难标记),求赐教

from awesome-ocr.

wanghaisheng avatar wanghaisheng commented on June 10, 2024

@violet-day 我理解你想问的是自然场景下的文本检测 你可以在看下这几个 大体上有针对word的 也有针对整行的

检测单词,而不是检测出一个文本行

from awesome-ocr.

violet-day avatar violet-day commented on June 10, 2024

@wanghaisheng 可能是请教的问题有点模糊,比如我现在是基于line去选择算法,是否一定要将样本中的所有文本行都标注出来,比如上面的营业范围不在业务要求识别的,在做样本标注时,如果忽略不标注对训练是否影响

from awesome-ocr.

wanghaisheng avatar wanghaisheng commented on June 10, 2024

@violet-day 你这个问题我由于没有实操过定位这部分 回答不了你
我理解应该没有很大影响
你有没有测试上面别人已经训练好的ctpn的检测

from awesome-ocr.

violet-day avatar violet-day commented on June 10, 2024

@wanghaisheng 用ctpn训练好的模型测试了一下,效果还可以,找时间研究下,谢谢楼主

from awesome-ocr.

wanghaisheng avatar wanghaisheng commented on June 10, 2024

@violet-day 客气 具体训练数据的准备 调优建议咨询一下 https://github.com/eragonruan/text-detection-ctpn

from awesome-ocr.

Sanster avatar Sanster commented on June 10, 2024

感谢楼主的分享!

需要生成 OCR 识别训练数据的同学,可以参考下我的这个项目:https://github.com/Sanster/text_renderer

from awesome-ocr.

Mundanee avatar Mundanee commented on June 10, 2024

感谢楼主的分享,想请问一下,对于单字分割的识别,有没有比较好的方式呢,比如说银行卡身份证这类的规整文字区域,定位roi是非常容易的,但因为图像质量和反光角度的问题,经常会有些字难以切割,一直都没有想到好的解决办法。

from awesome-ocr.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.