@ -1,110 +1,107 @@
## HyperLPR3 高性能开源中文车牌识别框架
## HyperLPR3 - High Performance License Plate Recognition Framework.
#### [](https://pypi.org/project/hyperlpr3/)[](https://pypi.org/manage/project/hyperlpr3/releases/)
#### [](https://pypi.org/project/hyperlpr3/)[](https://pypi.org/manage/project/hyperlpr3/releases/)
[中文文档 ](README_CH.md )
### 一键安装
### Fast Install
`python -m pip install hyperlpr3`
###### 支持python3, 支持Windows Mac Linux 树莓派等 。
###### support： 。
###### 720p cpu real-time (st on MBP r15 2.2GHz haswell).
#### 快速体验
#### Fast Test
```bash
# 使用命令行测试 - 图像url地址
# image url
lpr3 sample -src https://koss.iyong.com/swift/v1/iyong_public/iyong_2596631159095872/image/20190221/1550713902741045679.jpg
# 使用命令行测试 - 本地图像路径
# image path
lpr3 sample -src images/test_img.jpg -det high
```
#### 快速上手
#### Fast Use
```python
# 导入opencv库
# import opencv
import cv2
# 导入依赖包
# import hyperlpr3
import hyperlpr3 as lpr3
# 实例化识别对象
# Instantiate object
catcher = lpr3.LicensePlateCatcher()
# 读取图片
# load image
image = cv2.imread("images/test_img.jpg")
# 识别结果
# print result
print(catcher(image))
```
#### 启动WebApi服务
#### Start the WebAPI service
```bash
# 启动服务
# start server
lpr3 rest --port 8715 --host 0.0.0.0
```
启动后可打开SwaggerUI的路径： :
Path to open SwaggerUI after startup： :
#### Q& A
Q： Android识别率没有所传demo apk的识别率高 ？
Q： Whether the accuracy of android in the project is consistent with that of apk-demo ？
A： 请自行编译或从release中下载安卓动态库放置于Prj-Android中进行测试 。
A： Please compile or download the Android shared library from the release and copy it to Prj-Android for testing 。
Q： 车牌的训练数据来源 ？
Q： Source of training data for license plates ？
A： 由于用于训练车牌数据涉及到法律隐私等问题，本项目无法提供。开放较为大的数据集有[CCPD](https://github.com/detectRecog/CCPD)车牌数据集 。
A： Since the license plate data used for training involves legal privacy and other issues, it cannot be provided in this project. Open more big data sets [CCPD ](https://github.com/detectRecog/CCPD ) registration dataset 。
Q： 训练代码的提供 ？
Q： Provision of training code ？
A： 相关资源中有提供老版的训练代码， 。
A： The resources provide the old training code, and the training methods for HyperLPR3 will be sorted out and presented later 。
Q： ？
A： ， ，
#### 相关资源
#### Resources
- [五分钟搞定: 中文车牌识别光速部署与使用 ](https://blog.csdn.net/weixin_40193776/article/details/129258107 )
- [Linux/MacOS使用： ](https://blog.csdn.net/weixin_40193776/article/details/129295679 )
- 待补充...欢迎投稿
- To be added... Contributions welcome
#### 其他版本
#### Other versions
- [HyperLPRv1 版 ](https://github.com/szad670401/HyperLPR/tree/v1 )
- [HyperLPRv1 ](https://github.com/szad670401/HyperLPR/tree/v1 )
- [HyperLPRv2 版 ](https://github.com/szad670401/HyperLPR/tree/v2 )
- [HyperLPRv2 ](https://github.com/szad670401/HyperLPR/tree/v2 )
### TODO
- 支持快速部署WebApi服务
- 支持多种车牌以及双层
- 支持大角度车牌
- 轻量级识别模型
- Support for rapid deployment of WebApi services
- Support multiple license plates and double layers
- Support large Angle license plate
- Lightweight recognition model
### 特性
### Specialty
- 速度快 720p,单核 Intel 2.2G CPU (MaBook Pro 2015)平均识别时间低于 100ms
- 基于端到端的车牌识别无需进行字符分割
- 识别率高,卡口场景准确率在95%-97%左右
- 支持跨平台编译和快速部署
- 720p faster, single core Intel 2.2G CPU (MaBook Pro 2015) average recognition time is less than 100ms
- End-to-end license plate recognition does not require character segmentation
- The recognition rate is high, and the accuracy of the entrance and exit scene is about 95%-97%
- Support cross-platform compilation and rapid deployment
### 注意事项 :
### Matters Need Attention :
- 本项目的C++实现和Python实现无任何关联，
- 在编译C++工程的时候必须要使用OpenCV 4.0和MNN 2.0以上版本，否则无法编译
- 安卓工程编译ndk尽量采用21版本
- The C++ and Python implementations of this project are separate
- When compiling C++ projects, OpenCV 4.0 and MNN 2.0 must be used, otherwise it will not compile
- Android project compilation ndk as far as possible to use version 21
### Python 依赖
### Python Dependency
- opencv-python (>3.3)
- onnxruntime (>1.8.1)
- fastapi (0.92.0)
@ -114,99 +111,99 @@ A： ，
- tqdm
- requests
### 跨平台支持
### Cross-platform support
#### 平台
#### Platform
- Linux: x86、Armv7、Armv8
- MacOS: x86
- Android: arm64-v8a、armeabi-v7a
#### 开发板
#### Embedded Development Board
- Rockchip: rv1109rv1126(RKNPU)
### CPP 依赖
### CPP Dependency
- Opencv 4.0 以上版本
- MNN 2.0 以上版本
- Opencv 4.0 above
- MNN 2.0 above
### C/C++编译依赖库
### C/C++ Compiling dependencies
编译C/C++工程需要使用第三方依赖库，将库下载后解压，并将其通过拷贝或软链接放入根目录(与CMakeLists.txt同级)即可，依赖的库下载地址：[百度网盘 ](https://pan.baidu.com/s/138O2bSlPN0H81OYP6zc3yQ) code: 5duf
Compiling C/C++ projects requires the use of third-party dependency libraries. After downloading the library, unzip it, and put it into the root directory (the same level as CMakeLists.txt) by copying or soft linking.[baidu drive ](https://pan.baidu.com/s/138O2bSlPN0H81OYP6zc3yQ) code: 5duf
### Linux/Mac动态链接库编译
### Linux/Mac Shared Library Compilation
- 需要将依赖库放置或链接在项目根目录下(与CMakeLists.txt同级 )
- Need to place or link dependencies in the project root (same level as CMakeLists.txt )
```bash
# 执行编译脚本
# execute the script
sh command/build_release_linux_share.sh
```
编译后的相关物料放置于根目录下**build/linux/install/hyperlpr3**中，其中包含 ：
- include 头文件
- lib 动态库路径
- resource 包含测试图片与模型等静态资源
Compiled to the **build/linux/install/hyperlpr3** dir， ：
- include - header file
- lib - shared dir
- resource - test-images and models dir
按需取走需要的文件即可
Copy the files you need into your project
### Linux/Mac编译 Demo
### Linux/Mac Compiling the Demo
- 需要完成上一步的编译动作，并保证编译成功且编译完成后的物料放置于根目录下的**build/linux/install/hyperlpr3**路径中
- 需要从根目录中进入到子工程**Prj-Linux**文件夹中进行操作
- You need to complete the compilation action of the previous step，
- Go to the **Prj-Linux** folder
```bash
# 进入到子工程demo
# go to Prj-linux
cd Prj-Linux
# 创建build文件夹并进入
# make build and enter
mkdir build & & cd build
# 开始编译
# Start compiling
cmake .. & & make -j
```
编译完成后生成可执行程序**PlateRecDemo**，执行运行测试
The executable program is generated after compilation: **PlateRecDemo** ，
```bash
# 传入模型文件夹路径和需要预测的图像执行程序
# first param models dir, second param image path
./PlateRecDemo ../hyperlpr3/resource/models/r2_mobile ../hyperlpr3/resource/images/test_img.jpg
```
### Linux/Mac快速使用SDK代码示例
### Linux/Mac Quick use SDK code example
```C
// 读取图像
// Load image
cv::Mat image = cv::imread(image_path);
// 创建 ImageData
// Create a ImageData
HLPR_ImageData data = {0};
data.data = image.ptr< uint8_t > (0); // 设置图像数据流
data.width = image.cols; // 设置图像宽
data.height = image.rows; // 设置图像高
data.format = STREAM_BGR; // 设置当前图像编码格式
data.rotation = CAMERA_ROTATION_0; // 设置当前图像转角
// 创建数据 Buffer
data.data = image.ptr< uint8_t > (0); // Setting the image data flow
data.width = image.cols; // Setting the image width
data.height = image.rows; // Setting the image height
data.format = STREAM_BGR; // Setting the current image encoding format
data.rotation = CAMERA_ROTATION_0; // Setting the current image corner
// Create a Buffer
P_HLPR_DataBuffer buffer = HLPR_CreateDataBuffer(&data);
// 配置车牌识别参数
// Configure license plate recognition parameters
HLPR_ContextConfiguration configuration = {0};
configuration.models_path = model_path; // 模型文件夹路径
configuration.max_num = 5; // 最大识别车牌数量
configuration.det_level = DETECT_LEVEL_LOW; // 检测器等级
configuration.models_path = model_path; // Model folder path
configuration.max_num = 5; // Maximum number of license plates
configuration.det_level = DETECT_LEVEL_LOW; // Level of detector
configuration.use_half = false;
configuration.nms_threshold = 0.5f; // 非极大值抑制置信度阈值
configuration.rec_confidence_threshold = 0.5f; // 车牌号文本阈值
configuration.box_conf_threshold = 0.30f; // 检测器阈值
configuration.nms_threshold = 0.5f; // Non-maxima suppress the confidence threshold
configuration.rec_confidence_threshold = 0.5f; // License plate number text threshold
configuration.box_conf_threshold = 0.30f; // Detector threshold
configuration.threads = 1;
// 实例化车牌识别算法 Context
// Instantiating a Context
P_HLPR_Context ctx = HLPR_CreateContext(&configuration);
// 查询实例化状态
// Query the Context state
HREESULT ret = HLPR_ContextQueryStatus(ctx);
if (ret != HResultCode::Ok) {
printf("create error.\n");
return -1;
}
HLPR_PlateResultList results = {0};
// 执行车牌识别算法
// Execute LPR
HLPR_ContextUpdateStream(ctx, buffer, &results);
for (int i = 0; i < results.plate_size ; + + i ) {
// 解析识别后的数据
// Getting results
std::string type;
if (results.plates[i].type == HLPR_PlateType::PLATE_TYPE_UNKNOWN) {
type = "未知 ";
type = “Unknown ";
} else {
type = TYPES[results.plates[i].type];
}
@ -215,27 +212,27 @@ for (int i = 0; i < results.plate_size; ++i) {
results.plates[i].code, results.plates[i].text_confidence);
}
// 销毁 Buffer
// Release Buffer
HLPR_ReleaseDataBuffer(buffer);
// 销毁 Context
// Release Context
HLPR_ReleaseContext(ctx);
```
### Android编译动态链接库
- 需要完成上一步的编译动作，并保证编译成功且编译完成后的物料放置于根目录下的**build/linux/install/hyperlpr3**路径中
### Android: Compile the Shared Library
- The first step is to install third-party dependencies
```bash
# 执行编译脚本
# execute the script
sh command/build_release_android_share.sh
```
编译完成后android的动态库会放置于**build/release_android/**，其中包含 ：
- arm64-v8a 64位的动态库
- armeabi-v7a 32位的动态库
Compiled to the: **build/release_android/** ， ：
- arm64-v8a - 64bit shard library
- armeabi-v7a - 32bit shard library
完成Android的动态库编译后， ，
After compiling， **Prj-Android/hyperlpr3/libs** ， **Prj-Android** project to use.
###
### 可识别和待支持的车牌的类型
### License plate type is supported(Chinese)
#### 支持
- [x] 单行蓝牌
@ -253,15 +250,16 @@ sh command/build_release_android_share.sh
- [ ] 双层军牌
- [ ] 双层农用车牌
- [ ] 双层个性化车牌
- [ ] License plates from more countries
###### Note:由于训练的时候样本存在一些不均衡的问题,一些特殊车牌存在一定识别率低下的问题，如(使馆/港澳车牌)，会在后续的版本进行改进。
###### Note:Due to some imbalanced samples during training, some special license plates have low recognition rates, such as (Embassy/Hong Kong and Macao license plates), which will be improved in the subsequent versions.
### 识别测试APP
### Demo APP Install
- 体验 Android APP： ](http://fir.tunm.top/hyperlpr)
- Android APP： ](http://fir.tunm.top/hyperlpr)
#### 获取帮助
#### Help
- HyperAI讨论QQ群: 529385694，
- HyperAI QQ Group: 529385694
tunmx
2 years ago