五、OCR-PaddlePaddle训练源码解析系列-文字检测

一、linux调试

1、解压创建时添加的数据集

unzip -d ./PaddleOCR/train_data data/data158454/train_data.zip
1

2、下载模型到特定目录

wget -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/pretrained/MobileNetV3_large_x0_5_pretrained.pdparams
1

2.1、解释

wget命令详解：https://www.cnblogs.com/sx66/p/11887022.html

-P,  --directory-prefix=PREFIX  以 PREFIX/... 保存文件
1

3、安装包

pip install -r requirements.txt
1

4、训练

cd  PaddleOCR
1

python3 tools/train.py -c configs/det/det_mv3_db.yml \
     -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained

1
2
3

4.1、报错

AttributeError: module 'cv2' has no attribute '_registerMatType'
1

4.2、解决方案

There is mess with versions:

_registerMatType is a part of OpenCV 4.5.4
but you mention 4.5.1
Please ensure that you have installed one version of OpenCV only.
1
2
3
4
5

pip install opencv-python --upgrade
1

二、win10调试

1、训练train

python3 train.py -c ../configs/det/det_mv3_db.yml 
-o Global.pretrained_model=../pretrain_models/MobileNetV3_large_x0_5_pretrained Global.use_gpu=false Train.dataset.data_dir=../train_data/icdar2015/text_localization/ Train.dataset.label_file_list=[../train_data/icdar2015/text_localization/train_icdar2015_label.txt]
Train.loader.batch_size_per_card=2
Eval.dataset.data_dir=../train_data/icdar2015/text_localization/ Eval.dataset.label_file_list=[../train_data/icdar2015/text_localization/train_icdar2015_label.txt]
1
2
3
4

1.1、运行报错

ModuleNotFoundError: No module named ‘yaml‘
1

1.2、解决方案

 pip install pyyaml
1

1.3、配置理解

-代表后面的是列表，列表里面的元素可以是字典，也可以是值本身
在这里插入图片描述

1.4、yaml类型转换python类型

config = yaml.load(open(file_path, 'rb'), Loader=yaml.Loader)
1

1.5、python类型转换yaml类型

 with open(os.path.join(save_model_dir, 'config.yml'), 'w') as f:
            yaml.dump(
                dict(config), f, default_flow_style=False, sort_keys=False)
1
2
3

1.6、eval() 函数

dataset = eval(module_name)(config, mode, logger, seed)
1

eval() 函数用来执行一个字符串表达式，并返回表达式的值。

语法
以下是 eval() 方法的语法:

eval(expression[, globals[, locals]])
参数
expression – 表达式。
globals – 变量作用域，全局命名空间，如果被提供，则必须是一个字典对象。
locals – 变量作用域，局部命名空间，如果被提供，可以是任何映射对象。
返回值
返回表达式计算结果。

1.7、sample(self, population, k)

"""Chooses k unique random elements from a population sequence or set.
1

1.8、 eval（）字符串配置转换成op类

def create_operators(op_param_list, global_config=None):
    """
    create operators based on the config

    Args:
        params(list): a dict list, used to create some operators
    """
    assert isinstance(op_param_list, list), ('operator config should be a list')
    ops = []
    for operator in op_param_list:
        assert isinstance(operator,
                          dict) and len(operator) == 1, "yaml format error"
        op_name = list(operator)[0]
        param = {} if operator[op_name] is None else operator[op_name]
        if global_config is not None:
            param.update(global_config)
        op = eval(op_name)(**param)
        ops.append(op)
    return ops

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

1.9、输入数据和label的特征提取和转换

def transform(data, ops=None):
    """ transform """
    if ops is None:
        ops = []
    for op in ops:
        data = op(data)
        if data is None:
            return None
    return data
1
2
3
4
5
6
7
8
9