README.md 4.39 KB
Newer Older
liucong's avatar
liucong committed
1
# PaddleOCR
shizhm's avatar
shizhm committed
2

liucong's avatar
liucong committed
3
## 论文
4

liucong's avatar
liucong committed
5
6
7
8
9
10
11
12
13
PaddleOCR通过det、rec、cls三个模型分别实现字符检测、字符识别和字符方向分类的应用

det模型主要用DB算法,参考论文如下:

https://arxiv.org/pdf/1911.08947.pdf

rec模型主要用SVTR算法,参考论文如下:

https://arxiv.org/pdf/2205.00159.pdf
14
15
16

## 模型结构

liucong's avatar
liucong committed
17
det:
18

liucong's avatar
liucong committed
19
<img src="./Doc/Images/dbnet-arc.png" style="zoom:100%;" align=middle>
liucong's avatar
liucong committed
20

liucong's avatar
liucong committed
21
rec:
shizhm's avatar
shizhm committed
22

liucong's avatar
liucong committed
23
<img src="./Doc/Images/SVTR-arc.png" style="zoom:100%;" align=middle>
24

liucong's avatar
liucong committed
25
## 算法原理
26

liucong's avatar
liucong committed
27
百度PaddleOCR开源项目提供了车牌识别的预训练模型,本示例使用PaddleOCR提供的蓝绿黄牌识别模型进行推理。其中,DBnet是一种基于分割的文本检测方法,相比传统分割方法需要设定固定阈值,该模型将二值化操作插入到分割网络中进行联合优化,通过网络学习可以自适应的预测图像中每一个像素点的阈值,能够在像素水平很好的检测自然场景下不同形状的文字。SVTR是一种端到端的文本识别模型,通过单个视觉模型就可以一站式解决特征提取和文本转录两个任务,同时也保证了更快的推理速度。
28

liucong's avatar
liucong committed
29
30
<img src="./Doc/Images/paddleocr.png" style="zoom:100%;" align=middle>

liucong's avatar
liucong committed
31
32
## 环境配置

liucong's avatar
liucong committed
33
### Docker(方法一)
liucong's avatar
liucong committed
34
35

拉取镜像:
liucong's avatar
liucong committed
36

shizhm's avatar
shizhm committed
37
```
liucong's avatar
liucong committed
38
docker pull image.sourcefind.cn:5000/dcu/admin/base/migraphx:4.3.0-ubuntu20.04-dtk24.04.1-py3.10
shizhm's avatar
shizhm committed
39
40
```

liucong's avatar
liucong committed
41
创建并启动容器:
liucong's avatar
liucong committed
42
43

```
liucong's avatar
liucong committed
44
docker run --shm-size 16g --network=host --name=paddleocr_migraphx --privileged --device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v $PWD/paddleocr_migraphx:/home/paddleocr_migraphx -v /opt/hyhal:/opt/hyhal:ro -it <Your Image ID> /bin/bash
liucong's avatar
liucong committed
45

liucong's avatar
liucong committed
46
47
# 激活dtk
source /opt/dtk/env.sh
liucong's avatar
liucong committed
48
49
```

liucong's avatar
liucong committed
50
51
52
53
54
55
### Dockerfile(方法二)

```
cd ./docker
docker build --no-cache -t paddleocr_migraphx:2.0 .

liucong's avatar
liucong committed
56
docker run --shm-size 16g --network=host --name=paddleocr_migraphx --privileged --device=/dev/kfd --device=/dev/dri --group-add video --cap-add=SYS_PTRACE --security-opt seccomp=unconfined -v $PWD/paddleocr_migraphx:/home/paddleocr_migraphx -v /opt/hyhal:/opt/hyhal:ro -it <Your Image ID> /bin/bash
liucong's avatar
liucong committed
57
58
59

# 激活dtk
source /opt/dtk/env.sh
liucong's avatar
liucong committed
60
61
```

liucong's avatar
liucong committed
62
## 数据集
shizhm's avatar
shizhm committed
63

liucong's avatar
liucong committed
64
根据输入的样本图像,进行车牌识别。
liucong's avatar
liucong committed
65

liucong's avatar
liucong committed
66
67
68
69
70
## 推理

### Python版本推理

下面介绍如何运行Python代码示例,Python示例的详细说明见Doc目录下的Tutorial_Python.md。
liucong's avatar
liucong committed
71

liucong's avatar
liucong committed
72
#### 设置环境变量
liucong's avatar
liucong committed
73
74

```
liucong's avatar
liucong committed
75
export PYTHONPATH=/opt/dtk/lib:$PYTHONPATH
liucong's avatar
liucong committed
76
77
```

liucong's avatar
liucong committed
78
#### 运行示例
liucong's avatar
liucong committed
79

liucong's avatar
liucong committed
80
81
82
```
# 进入python示例目录
cd <path_to_paddleocr_migraphx>/Python
liucong's avatar
liucong committed
83

liucong's avatar
liucong committed
84
85
# 安装依赖
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
shizhm's avatar
shizhm committed
86

liucong's avatar
liucong committed
87
88
# 运行示例
python PaddleOCR_infer_migraphx.py
shizhm's avatar
shizhm committed
89
```
liucong's avatar
liucong committed
90

liucong's avatar
liucong committed
91
### C++版本推理
Your Name's avatar
Your Name committed
92

liucong's avatar
liucong committed
93
94
95
96
97
98
99
100
101
102
注意:当使用操作系统不一样时,CMakeList需要做相应的修改:

```
# ubuntu操作系统
${CMAKE_CURRENT_SOURCE_DIR}/depend/lib64/ 修改为 ${CMAKE_CURRENT_SOURCE_DIR}/depend/lib/

# centos操作系统
${CMAKE_CURRENT_SOURCE_DIR}/depend/lib/ 修改为 ${CMAKE_CURRENT_SOURCE_DIR}/depend/lib64/
```

liucong's avatar
liucong committed
103
104
105
106
下面介绍如何运行C++代码示例,C++示例的详细说明见Doc目录下的Tutorial_Cpp.md。


#### 构建工程
Your Name's avatar
Your Name committed
107
108
109

```
rbuild build -d depend
110
111
```

liucong's avatar
liucong committed
112
#### 设置环境变量
Your Name's avatar
Your Name committed
113
114
115

将依赖库依赖加入环境变量LD_LIBRARY_PATH,在~/.bashrc中添加如下语句:

liucong's avatar
liucong committed
116
117
118
当操作系统是ubuntu系统时:

```
liucong's avatar
liucong committed
119
export LD_LIBRARY_PATH=<path_to_paddleocr_migraphx>/depend/lib/:$LD_LIBRARY_PATH
liucong's avatar
liucong committed
120
121
122
123
```

当操作系统是centos系统时:

124
```
shizhm's avatar
shizhm committed
125
export LD_LIBRARY_PATH=<path_to_paddleocr_migraphx>/depend/lib64/:$LD_LIBRARY_PATH
Your Name's avatar
Your Name committed
126
127
128
```

然后执行:
129

Your Name's avatar
Your Name committed
130
131
132
133
```
source ~/.bashrc
```

liucong's avatar
liucong committed
134
#### 运行示例
Your Name's avatar
Your Name committed
135

liucong's avatar
liucong committed
136
成功编译PaddleOCR车牌识别工程后,执行如下命令运行该示例:
137
138

```
shizhm's avatar
shizhm committed
139
# 进入paddleocr migraphx工程根目录
shizhm's avatar
shizhm committed
140
cd <path_to_paddleocr_migraphx> 
liucong's avatar
liucong committed
141
142
143
144

# 进入build目录
cd ./build/

145
# 运行示例
shizhm's avatar
shizhm committed
146
./PaddleOCR_VLPR
Your Name's avatar
Your Name committed
147
148
```

liucong's avatar
liucong committed
149
150
151
152
153
154
155
## result

#### Python版本

输入样本图像,进行车牌识别:

<img src="./Doc/Images/vlpr.jpg" style="zoom:100%;" align=middle>
Your Name's avatar
Your Name committed
156
157

```
shizhm's avatar
shizhm committed
158
皖AD19906
Your Name's avatar
Your Name committed
159
160
```

liucong's avatar
liucong committed
161
162
163
164
165
166
167
168
169
170
#### C++版本

输入样本图像,进行车牌识别:

<img src="./Doc/Images/vlpr.jpg" style="zoom:100%;" align=middle>

```
皖AD19906
```

liucong's avatar
liucong committed
171
172
173
174
### 精度



liucong's avatar
liucong committed
175
176
177
178
179
180
181
182
## 应用场景

### 算法类别

`ocr`

### 热点应用行业

liucong's avatar
liucong committed
183
`金融`,`交通`,`教育`,`医疗`
liucong's avatar
liucong committed
184

shizhm's avatar
shizhm committed
185
## 源码仓库及问题反馈
186

chenzk's avatar
chenzk committed
187
https://developer.sourcefind.cn/codes/modelzoo/paddleocr_migraphx
188

liucong's avatar
liucong committed
189
## 参考资料
190

liucong's avatar
liucong committed
191
https://github.com/PaddlePaddle/PaddleOCR