FORMAT_CONVERSION_MAXCOMPUTE/ocr/ocr_interface.py

import base64
import json
import multiprocessing as mp
import sys
import os
sys.path.append(os.path.dirname(os.path.abspath(__file__)) + "/../")
import time
import traceback
from multiprocessing.context import Process
import cv2
import requests
import logging
import numpy as np
os.environ['FLAGS_eager_delete_tensor_gb'] = '0'
from ocr.paddleocr import PaddleOCR


logging.basicConfig(level=logging.INFO,format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)
def log(msg):
    '''
    @summary:打印信息
    '''
    logger.info(msg)


def ocr(data, ocr_model):
    logging.info("into ocr_interface ocr")
    try:
        img_data = base64.b64decode(data)
        text = picture2text(img_data, ocr_model)
        return text
    except TimeoutError:
        raise TimeoutError


flag = 0
def picture2text(img_data, ocr_model):
    logging.info("into ocr_interface picture2text")
    try:
        start_time = time.time()
        # 二进制数据流转np.ndarray [np.uint8: 8位像素]
        img = cv2.imdecode(np.frombuffer(img_data, np.uint8), cv2.IMREAD_COLOR)
        # 将bgr转为rbg
        try:
            np_images = cv2.cvtColor(img, cv2.COLOR_RGB2BGR)
        except cv2.error as e:
            if "src.empty()" in str(e):
                logging.info("ocr_interface picture2text image is empty!")
                return {"text": str([]), "bbox": str([])}
        # resize
        # cv2.imshow("before resize", np_images)
        # print("np_images.shape", np_images.shape)

        # best_h, best_w = get_best_predict_size(np_images)
        # np_images = cv2.resize(np_images, (best_w, best_h), interpolation=cv2.INTER_AREA)

        # cv2.imshow("after resize", np_images)
        # print("np_images.shape", np_images.shape)
        # cv2.waitKey(0)

        # 预测
        results = ocr_model.ocr(np_images, det=True, rec=True, cls=True)

        # 循环每张图片识别结果
        text_list = []
        bbox_list = []
        for line in results:
            # print("ocr_interface line", line)
            text_list.append(line[-1][0])
            bbox_list.append(line[0])

        # 查看bbox
        # img = np.zeros((np_images.shape[1], np_images.shape[0]), np.uint8)
        # img.fill(255)
        # for box in bbox_list:
        #     print(box)
        #     cv2.rectangle(img, (int(box[0][0]), int(box[0][1])),
        #                   (int(box[2][0]), int(box[2][1])), (0, 0, 255), 1)
        # cv2.imshow("bbox", img)
        # cv2.waitKey(0)

        logging.info("ocr model use time: " + str(time.time()-start_time))
        return {"text": str(text_list), "bbox": str(bbox_list)}

    except TimeoutError:
        raise TimeoutError
    except Exception as e:
        logging.info("picture2text error!")
        print("picture2text", traceback.print_exc())
        return {"text": str([]), "bbox": str([])}


def get_best_predict_size(image_np):
    sizes = [1280, 1152, 1024, 896, 768, 640, 512, 384, 256, 128]

    min_len = 10000
    best_height = sizes[0]
    for height in sizes:
        if abs(image_np.shape[0] - height) < min_len:
            min_len = abs(image_np.shape[0] - height)
            best_height = height

    min_len = 10000
    best_width = sizes[0]
    for width in sizes:
        if abs(image_np.shape[1] - width) < min_len:
            min_len = abs(image_np.shape[1] - width)
            best_width = width

    return best_height, best_width


class OcrModels:
    def __init__(self):
        try:
            self.ocr_model = PaddleOCR(use_angle_cls=True, lang="ch")
        except:
            print(traceback.print_exc())
            raise RuntimeError

    def get_model(self):
        return self.ocr_model


if __name__ == '__main__':
    # if len(sys.argv) == 2:
    #     port = int(sys.argv[1])
    # else:
    #     port = 15011
    #
    # app.run(host='0.0.0.0', port=port, threaded=False, debug=False)
    # log("OCR running")
    file_path = "C:/Users/Administrator/Desktop/error1.png"
    # file_path = "1.png"

    with open(file_path, "rb") as f:
        file_bytes = f.read()
    file_base64 = base64.b64encode(file_bytes)

    ocr_model = OcrModels().get_model()
    result = ocr(file_base64, ocr_model)
    result = ocr(file_base64, ocr_model)

    text_list = eval(result.get("text"))
    box_list = eval(result.get("bbox"))

    new_list = []
    for i in range(len(text_list)):
        new_list.append([text_list[i], box_list[i]])

    # print(new_list[0][1])
    new_list.sort(key=lambda x: (x[1][1][0], x[1][0][0]))

    for t in new_list:
        print(t[0])