基于百度飞桨PaddleOCR的图片文字识别

PaddleOCR项目源码：https://github.com/PaddlePaddle/PaddleOCR

飞桨开源文字识别模型套件PaddleOCR，目标是打造丰富、领先、实用的文本识别模型/工具库。最新开源的超轻量PP-OCRv3模型大小仅为16.2M。同时支持中英文识别；支持倾斜、竖排等多种方向文字识别；支持GPU、CPU预测；用户既可以通过PaddleHub很便捷的直接使用该超轻量模型，也可以使用PaddleOCR开源套件训练自己的超轻量模型。

本文基于上述源码，先下下来，主要用来学习，使用PaddleOCR模型，进行测试。

一、编译环境
电脑上装了Anaconda3 + Pycharm + python3.8的环境。就基于此环境。
新建项目时，选择Anaconda3 下的python.exe(python3.8)

二、安装库
因为使用Anaconda3，直接通过其命令终端来安装，加上百度源，速度还是蛮快的。
1、安装安装PaddlePaddle

pip install paddlepaddle -i https://mirror.baidu.com/pypi/simple

2、安装PaddleOCR whl包

pip install "paddleocr>=2.0.1" # 推荐使用2.0.1+版本

安装paddleocr过程中，会出现一些模块安装失败的提示，可以先通过 pip install 模块，安装完了再安装paddleocr。

3、查看安装库的版本：

python -c "import paddle; print(paddle.__version__)"

三、新建项目
注意解释器的选择，见第一步。
1、将Github下载的源码PaddleOCR-release-2.6中的文件夹ppocr、ppstructure、tools拷贝到ocrtest文件夹下；

2、下载PaddleOCR提供的测试图片https://paddleocr.bj.bcebos.com/dygraph_v2.1/ppocr_img.zip，解压后，将fonts文件夹拷贝到ocrtest文件夹下；

3、ocrtest文件夹下新建img文件夹，里面存放“test.jpg”；test.jpg图片可以从2中下载的测试图片里找。

4、新建python文件paddleocr.py，项目结构如下：

代码如下：

# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import os
import sys
import importlib

__dir__ = os.path.dirname(__file__)

from types import ModuleType

import paddle

sys.path.append(os.path.join(__dir__, ''))

import cv2
import logging
import numpy as np
from pathlib import Path

tools = importlib.import_module('.', 'tools')
ppocr = importlib.import_module('.', 'ppocr')
ppstructure = importlib.import_module('.', 'ppstructure')

from tools.infer import predict_system
from ppocr.utils.logging import get_logger

logger = get_logger()
from ppocr.utils.utility import check_and_read, get_image_file_list
from ppocr.utils.network import maybe_download, download_with_progressbar, is_link, confirm_model_dir_url
from tools.infer.utility import draw_ocr, str2bool, check_gpu
from ppstructure.utility import init_args, draw_structure_result
from ppstructure.predict_system import StructureSystem, save_structure_res, to_excel


SUPPORT_DET_MODEL = ['DB']
VERSION = '2.6.1.0'
SUPPORT_REC_MODEL = ['CRNN', 'SVTR_LCNet']
BASE_DIR = os.path.expanduser("~/.paddleocr/")

DEFAULT_OCR_MODEL_VERSION = 'PP-OCRv3'
SUPPORT_OCR_MODEL_VERSION = ['PP-OCRv3']

MODEL_URLS = {
    'OCR': {
        'PP-OCRv3': {
            'det': {
                'ch': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_infer.tar',
                },
                'en': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/english/en_PP-OCRv3_det_infer.tar',
                },
                'ml': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/Multilingual_PP-OCRv3_det_infer.tar'
                }
            },
            'rec': {
                'ch': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/ppocr_keys_v1.txt'
                },
                'en': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/english/en_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/en_dict.txt'
                },
                'korean': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/korean_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/korean_dict.txt'
                },
                'japan': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/japan_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/japan_dict.txt'
                },
                'chinese_cht': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/chinese_cht_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/chinese_cht_dict.txt'
                },
                'ta': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/ta_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/ta_dict.txt'
                },
                'te': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/te_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/te_dict.txt'
                },
                'ka': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/ka_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/ka_dict.txt'
                },
                'latin': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/latin_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/latin_dict.txt'
                },
                'arabic': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/arabic_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/arabic_dict.txt'
                },
                'cyrillic': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/cyrillic_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/cyrillic_dict.txt'
                },
                'devanagari': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/PP-OCRv3/multilingual/devanagari_PP-OCRv3_rec_infer.tar',
                    'dict_path': './ppocr/utils/dict/devanagari_dict.txt'
                },
            },
            'cls': {
                'ch': {
                    'url':
                    'https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar',
                }
            },
        }
    }
}

def parse_args(mMain=True):
    import argparse
    parser = init_args()
    parser.add_help = mMain
    parser.add_argument("--lang", type=str, default='ch')
    parser.add_argument("--det", type=str2bool, default=True)
    parser.add_argument("--rec", type=str2bool, default=True)
    parser.add_argument("--type", type=str, default='ocr')
    parser.add_argument(
        "--ocr_version",
        type=str,
        choices=SUPPORT_OCR_MODEL_VERSION,
        default='PP-OCRv3',
        help='OCR Model version, the current model support list is as follows: '
        '1. PP-OCRv3 Support Chinese and English detection and recognition model, and direction classifier model'
        '2. PP-OCRv2 Support Chinese detection and recognition model. '
        '3. PP-OCR support Chinese detection, recognition and direction classifier and multilingual recognition model.'
    )

    for action in parser._actions:
        if action.dest in [
                'rec_char_dict_path', 'table_char_dict_path', 'layout_dict_path'
        ]:
            action.default = None
    if mMain:
        return parser.parse_args()
    else:
        inference_args_dict = {}
        for action in parser._actions:
            inference_args_dict[action.dest] = action.default
        return argparse.Namespace(**inference_args_dict)

def parse_lang(lang):
    latin_lang = [
        'af', 'az', 'bs', 'cs', 'cy', 'da', 'de', 'es', 'et', 'fr', 'ga', 'hr',
        'hu', 'id', 'is', 'it', 'ku', 'la', 'lt', 'lv', 'mi', 'ms', 'mt', 'nl',
        'no', 'oc', 'pi', 'pl', 'pt', 'ro', 'rs_latin', 'sk', 'sl', 'sq', 'sv',
        'sw', 'tl', 'tr', 'uz', 'vi', 'french', 'german'
    ]
    arabic_lang = ['ar', 'fa', 'ug', 'ur']
    cyrillic_lang = [
        'ru', 'rs_cyrillic', 'be', 'bg', 'uk', 'mn', 'abq', 'ady', 'kbd', 'ava',
        'dar', 'inh', 'che', 'lbe', 'lez', 'tab'
    ]
    devanagari_lang = [
        'hi', 'mr', 'ne', 'bh', 'mai', 'ang', 'bho', 'mah', 'sck', 'new', 'gom',
        'sa', 'bgc'
    ]
    if lang in latin_lang:
        lang = "latin"
    elif lang in arabic_lang:
        lang = "arabic"
    elif lang in cyrillic_lang:
        lang = "cyrillic"
    elif lang in devanagari_lang:
        lang = "devanagari"

    assert lang in MODEL_URLS['OCR'][DEFAULT_OCR_MODEL_VERSION][
        'rec'], 'param lang must in {}, but got {}'.format(
            MODEL_URLS['OCR'][DEFAULT_OCR_MODEL_VERSION]['rec'].keys(), lang)
    if lang == "ch":
        det_lang = "ch"
    elif lang == 'structure':
        det_lang = 'structure'
    elif lang in ["en", "latin"]:
        det_lang = "en"
    else:
        det_lang = "ml"
    return lang, det_lang

def get_model_config(type, version, model_type, lang):
    if type == 'OCR':
        DEFAULT_MODEL_VERSION = DEFAULT_OCR_MODEL_VERSION
    else:
        raise NotImplementedError

    model_urls = MODEL_URLS[type]
    if version not in model_urls:
        version = DEFAULT_MODEL_VERSION
    if model_type not in model_urls[version]:
        if model_type in model_urls[DEFAULT_MODEL_VERSION]:
            version = DEFAULT_MODEL_VERSION
        else:
            logger.error('{} models is not support, we only support {}'.format(
                model_type, model_urls[DEFAULT_MODEL_VERSION].keys()))
            sys.exit(-1)

    if lang not in model_urls[version][model_type]:
        if lang in model_urls[DEFAULT_MODEL_VERSION][model_type]:
            version = DEFAULT_MODEL_VERSION
        else:
            logger.error(
                'lang {} is not support, we only support {} for {} models'.
                format(lang, model_urls[DEFAULT_MODEL_VERSION][model_type].keys(
                ), model_type))
            sys.exit(-1)
    return model_urls[version][model_type][lang]


def img_decode(content: bytes):
    np_arr = np.frombuffer(content, dtype=np.uint8)
    return cv2.imdecode(np_arr, cv2.IMREAD_COLOR)


def check_img(img):
    if isinstance(img, bytes):
        img = img_decode(img)
    if isinstance(img, str):
        # download net image
        if is_link(img):
            download_with_progressbar(img, 'tmp.jpg')
            img = 'tmp.jpg'
        image_file = img
        img, flag_gif, flag_pdf = check_and_read(image_file)
        if not flag_gif and not flag_pdf:
            with open(image_file, 'rb') as f:
                img = img_decode(f.read())
        if img is None:
            logger.error("error in loading image:{}".format(image_file))
            return None
    if isinstance(img, np.ndarray) and len(img.shape) == 2:
        img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)

    return img

class PaddleOCR(predict_system.TextSystem):
    def __init__(self, **kwargs):
        """
        paddleocr package
        args:
            **kwargs: other params show in paddleocr --help
        """
        params = parse_args(mMain=False)
        params.__dict__.update(**kwargs)
        assert params.ocr_version in SUPPORT_OCR_MODEL_VERSION, "ocr_version must in {}, but get {}".format(
            SUPPORT_OCR_MODEL_VERSION, params.ocr_version)
        params.use_gpu = check_gpu(params.use_gpu)

        if not params.show_log:
            logger.setLevel(logging.INFO)
        self.use_angle_cls = params.use_angle_cls
        lang, det_lang = parse_lang(params.lang)

        # init model dir
        det_model_config = get_model_config('OCR', params.ocr_version, 'det',
                                            det_lang)
        params.det_model_dir, det_url = confirm_model_dir_url(
            params.det_model_dir,
            os.path.join(BASE_DIR, 'whl', 'det', det_lang),
            det_model_config['url'])
        rec_model_config = get_model_config('OCR', params.ocr_version, 'rec',
                                            lang)
        params.rec_model_dir, rec_url = confirm_model_dir_url(
            params.rec_model_dir,
            os.path.join(BASE_DIR, 'whl', 'rec', lang), rec_model_config['url'])
        cls_model_config = get_model_config('OCR', params.ocr_version, 'cls',
                                            'ch')
        params.cls_model_dir, cls_url = confirm_model_dir_url(
            params.cls_model_dir,
            os.path.join(BASE_DIR, 'whl', 'cls'), cls_model_config['url'])
        if params.ocr_version == 'PP-OCRv3':
            params.rec_image_shape = "3, 48, 320"
        else:
            params.rec_image_shape = "3, 32, 320"
        # download model if using paddle infer
        if not params.use_onnx:
            maybe_download(params.det_model_dir, det_url)
            maybe_download(params.rec_model_dir, rec_url)
            maybe_download(params.cls_model_dir, cls_url)

        if params.det_algorithm not in SUPPORT_DET_MODEL:
            logger.error('det_algorithm must in {}'.format(SUPPORT_DET_MODEL))
            sys.exit(0)
        if params.rec_algorithm not in SUPPORT_REC_MODEL:
            logger.error('rec_algorithm must in {}'.format(SUPPORT_REC_MODEL))
            sys.exit(0)

        if params.rec_char_dict_path is None:
            params.rec_char_dict_path = str(
                Path(__file__).parent / rec_model_config['dict_path'])

        logger.debug(params)
        # init det_model and rec_model
        super().__init__(params)
        self.page_num = params.page_num

    def ocr(self, img, det=True, rec=True, cls=True):
        """
        ocr with paddleocr
        args：
            img: img for ocr, support ndarray, img_path and list or ndarray
            det: use text detection or not. If false, only rec will be exec. Default is True
            rec: use text recognition or not. If false, only det will be exec. Default is True
            cls: use angle classifier or not. Default is True. If true, the text with rotation of 180 degrees can be recognized. If no text is rotated by 180 degrees, use cls=False to get better performance. Text with rotation of 90 or 270 degrees can be recognized even if cls=False.
        """
        assert isinstance(img, (np.ndarray, list, str, bytes))
        if isinstance(img, list) and det == True:
            logger.error('When input a list of images, det must be false')
            exit(0)
        if cls == True and self.use_angle_cls == False:
            logger.warning(
                'Since the angle classifier is not initialized, the angle classifier will not be uesd during the forward process'
            )

        img = check_img(img)
        # for infer pdf file
        if isinstance(img, list):
            if self.page_num > len(img) or self.page_num == 0:
                self.page_num = len(img)
            imgs = img[:self.page_num]
        else:
            imgs = [img]
        if det and rec:
            ocr_res = []
            for idx, img in enumerate(imgs):
                dt_boxes, rec_res, _ = self.__call__(img, cls)
                tmp_res = [[box.tolist(), res]
                           for box, res in zip(dt_boxes, rec_res)]
                ocr_res.append(tmp_res)
            return ocr_res
        elif det and not rec:
            ocr_res = []
            for idx, img in enumerate(imgs):
                dt_boxes, elapse = self.text_detector(img)
                tmp_res = [box.tolist() for box in dt_boxes]
                ocr_res.append(tmp_res)
            return ocr_res
        else:
            ocr_res = []
            cls_res = []
            for idx, img in enumerate(imgs):
                if not isinstance(img, list):
                    img = [img]
                if self.use_angle_cls and cls:
                    img, cls_res_tmp, elapse = self.text_classifier(img)
                    if not rec:
                        cls_res.append(cls_res_tmp)
                rec_res, elapse = self.text_recognizer(img)
                ocr_res.append(rec_res)
            if not rec:
                return cls_res
            return ocr_res







if __name__ == '__main__':
    args = parse_args(mMain=True)

    image_dir = 'img/test.jpg'
    print(image_dir)
    image_file_list = get_image_file_list(image_dir)
    if len(image_file_list) == 0:
        logger.error('no images find in {}'.format(image_dir))
        # return

    engine = PaddleOCR()

    for img_path in image_file_list:
        img_name = os.path.basename(img_path).split('.')[0]
        logger.info('{}{}{}'.format('*' * 10, img_path, '*' * 10))

        result = engine.ocr(img_path,
                            det=True,   #识别
                            rec=True,   #检测
                            cls=True)   #使用方向分类器识别180度旋转文字
        if result is not None:
            for idx in range(len(result)):
                res = result[idx]
                for line in res:
                    logger.info(line)
        else:
            print("result is none")

        # 显示结果
        from PIL import Image

        result = result[0]
        image = Image.open(img_path).convert('RGB')
        boxes = [line[0] for line in result]
        txts = [line[1][0] for line in result]
        scores = [line[1][1] for line in result]
        im_show = draw_ocr(image, boxes, txts, scores, font_path='./fonts/simfang.ttf')
        im_show = Image.fromarray(im_show)
        im_show.save('img/result.jpg')

5、编译调试
中间出现了一些模块未定义的提示，比如

安装相应的库解决：pip install lanms-neo
直到编译完成。

四、测试结果：
运行结果是一个list，每个item包含了文本框，文字和识别置信度

img文件夹中保存的result.jpg如下：

图片开头的“www.997788.com”识别成了“r88.997788.co11”，“中国收藏热线”没识别出来，“G11”识别成了“C11”，有污迹的时间“15:55”没识别出来。