Giter VIP home page Giter VIP logo

cptn-crnn's Introduction

cptn-crnn

从别人那里拿来的,稍微改了下,适合识别竖排书法,原文:

https://github.com/eragonruan/text-detection-ctpn

https://github.com/Sierkinhane/crnn_chinese_characters_rec

cptn

环境搭建(tqdm,opencv-python,Shapely,matplotlib,numpy,tensorflow-gpu or tensorflow,Cython,ipython 自行用pip3安装)

cd cptn/utils/bbox
sh make.sh

创建数据集

cd cptn/utils/prepare
python3 split_label.py(DATA_FOLDER 和 OUTPUT 改成自己的路径)

原始数据图片: Image text

csv标记内容为:

FileName x1 y1 x2 y2 x3 y3 x4 y4 text
img_calligraphy_70001_bg.jpg 72 53 96 53 96 358 72 358 黎沈昨骑托那缝丁聚侮篮海炭
img_calligraphy_70001_bg.jpg 46 53 70 53 70 394 46 394 缩蝇躁劣趋拴局伦绸启杭吭惯蛋仅
img_calligraphy_70001_bg.jpg 20 53 44 53 44 174 20 174 效射市关蝉

创建好的数据集图片(绿线为标记的坐标,输入ctpn的图片是不带绿线的): Image text

lable格式在 data/demo/img_calligraphy_70001_bg.txt中

训练

1.首先下载vgg16的模型
https://github.com/tensorflow/models/tree/1af55e018eebce03fb61bba9959a04672536107d/research/slim
,放到data目录下

2.运行 python3 main/train.py
可以直接下载我训练好的模型:
https://drive.google.com/open?id=1RwZb1HLG0vum-5RHZdSfqtDD2in_sNRD

crnn

环境搭建

安装 pytorch和warp-ctc
根据cuda版本选择pytorch的安装文件
pip3 install https://download.pytorch.org/whl/cu100/torch-1.0.1.post2-cp36-cp36m-linux_x86_64.whl
pip3 install torchvision
git clone https://github.com/SeanNaren/warp-ctc.git
cd warp-ctc;mkdir build; cd build;cmake ..;make
cd ../warp-ctc/pytorch_binding;python setup.py install
将 pytorch_binding 中生成的warpctc_pytorch文件夹copy到crnn下

制作训练集

1、python3 to_lmdb/tolmdb.py(按比例将训练集划分成train和val,生成方式一样,就是放到不同的目录)
2、根据自己的lable文件,生成 alphabets.py(就是将所有lable去重写进此文件)

训练(第一次训练,--model_path '')

python3 crnn_main.py --trainroot './data/' --valroot './data/val' --cuda --model_path './expr/crnn_Rec_done_35_2019-03-27.pth'

训练结果

我的训练模型(35次可能对训练集有些过拟合,提供了中间的几个结果模型,看哪一个更适合):
https://drive.google.com/open?id=1Ckz1j5ZXfNILh1ePJlYcDpE_PQ-tqML- 
训练结果:
Test loss: 0.178429, accuray: 0.997500
[35/300][24100/29790] Loss: 0.658935
[35/300][24200/29790] Loss: 0.534306
[35/300][24300/29790] Loss: 0.541349
[35/300][24400/29790] Loss: 0.475645

Image text : 兽亵播疒阌飨百怒逭纫

Image text : 蜂肉昆材摄

Image text : 不似周趋阙去

cptn-crnn's People

Contributors

hell-to-heaven avatar hwwu avatar

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.