D2C机器学习生成CRM表单 - 服务端部署

表单智能生成 & 编辑器

https://hub.dappwind.com/crm-maker/

https://hub.dappwind.com/crm-editor/

接口

https://api.dappwind.com/ai/generate/form

method: POST

body: file 要识别的表单图片

原理

前端代码自动生成-机器学习 模型训练

https://blog.dappwind.com/2020/05/13/index.html

前端代码自动生成-Opencv提取后模型分类

https://blog.dappwind.com/2020/05/14/index.html

前端代码自动生成-文字识别并关联元素

https://blog.dappwind.com/2020/05/15/index.html

前端代码自动生成 之 根据识别出的内容生成前端代码

https://blog.dappwind.com/2020/05/18/index.html

包安装

pip3 install flask
pip3 install opencv-python-headless
# pip3 install tensorflow 
# 如果出现 killed 无法安装 可以尝试在后面加上 --no-cache-dir
pip3 install tensorflow --no-cache-dir
pip3 install keras
pip3 install pillow
pip3 install scipy
pip3 install pytesseract
pip3 install 'elastic-apm[flask]'

文字识别库安装

ubuntu

sudo apt install tesseract-ocr
sudo apt install tesseract-ocr-chi-sim

mac

brew install tesseract

然后复制 https://github.com/tesseract-ocr/tessdata/raw/master/chi_sim.traineddata 到 /usr/local/Cellar/tesseract/4.1.1/share/tessdata

已训练的model

https://objectstorage.ap-tokyo-1.oraclecloud.com/n/nrucepepr744/b/bucket-20210409-2233/o/1589869679-20210409T141912Z-001.zip

下载后解压到 model/1589869679 即可运行

启动命令

lsof -i tcp:5000 | grep LISTEN | awk '{print $2}' | xargs kill -9

nohup python3 -u app.py > app.log 2>&1 &