pkjy / tiny-captcha-solver

A tiny simple out-of-the-box api bypass slide captcha and ocr captcha. using opencv and tesseract. self training tessdata.验证码识别服务，支持识别滑动验证码的缺口位置以及数字字母OCR识别内容。

tiny-captcha-solver

A tiny simple out-of-the-box api for slide captcha and ocr captcha. using opencv and tesseract. self training tessdata. 一个简易的验证码识别服务，支持数字、字母的OCR以及滑动验证码的缺口识别。

运行环境

• debian 9/ debian 10 /debian 11
• python 3.8+
• opencv
• tesseract 4.0+

特性

• 滑块验证码的缺口位置识别
• 简单数字或字母的OCR识别
• 自训练数据集，增加对某些字体的识别成功率

服务器部署

# 获取代码
git clone https://github.com/pkjy/tiny-captcha-solver.git

# 进入项目目录
cd tiny-captcha-solver

# 安装依赖
apt-get install -y tesseract-ocr python3-opencv
pip3 install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/

# 拷贝识别的数据集
mkdir /tessdata
cp -r tessdata/* /

# 运行
python app.py

Docker部署

使用 Dockerfile 构建 或者直接 Pull镜像

# 获取代码
git clone https://github.com/pkjy/tiny-captcha-solver.git

# 进入项目目录
cd tiny-captcha-solver

# dockerfile 构建
docker build -t tiny-captcha-solver:latest .

# 运行镜像
docker run -p 5000:5000 --name tiny-captcha-solver -d tiny-captcha-solver:latest

# 从 dockerhub pull
docker pull pkjy/tiny-captcha-solver:latest
# 运行镜像
docker run -itd --rm -p 5000:5000 --name tiny-captcha-solver pkjy/tiny-captcha-solver:latest

use

send request in local, for slide.

POST 127.0.0.1:5000/slide/base64
content-type: application/json

# post data with body raw (json):
{
  target: base64 format for target image
  template: base64 format for full background image
}

will return target position {x1,x2,y1,y2} like

{
    "code": 0,
    "result": {
        "x1": "181",
        "x2": "249",
        "y1": "78",
        "y2": "146"
    }
}

raw input

result

send request in local, for ocr.

raw input

available arguments：

name	arguments	note
type	pkjy.num、pkjy.alphabet_num	default is pkjy.alphabet_num. num is for pure number ocr. alphabet_num is for combination of alphabet and number

POST 127.0.0.1:5000/ocr/base64
content-type: application/json

# post data with body raw (json):
{
  base64: [multi base64 format for target image] 
}

will return target position {"code": 0,"result": ["47SS"]} like

{
    "code": 0,
    "result": [
        "47SS"
    ]
}

demo

demo url: https://pkjy.xyz/captcha .you can request here with your images

notice

if you get src like data:image/jpg;base64,UklGRkgJAABXRUJQVlA4WAoAAAAQAAAAQwAAQwAAQUxQS... you need to remove header like data:image/jpg;base64, since it's Data URLs usage not base64 standard format.