1
from paddleocr import PaddleOCR, draw_ocr
2

3
# Paddleocr目前支持中英文、英文、法语、德语、韩语、日语,可以通过修改lang参数进行切换
4
# 参数依次为`ch`, `en`, `french`, `german`, `korean`, `japan`。
5
ocr = PaddleOCR(use_angle_cls=True, lang="ch")  # need to run only once to download and load model into memory
6
img_path = 'PaddleOCR/doc/imgs/11.jpg'
7
result = ocr.ocr(img_path, cls=True)
8
for idx in range(len(result)):
9
    res = result[idx]
10
    for line in res:
11
        print(line)
12

13
# 显示结果
14
from PIL import Image
15
result = result[0]
16
image = Image.open(img_path).convert('RGB')
17
boxes = [line[0] for line in result]
18
txts = [line[1][0] for line in result]
19
scores = [line[1][1] for line in result]
20
im_show = draw_ocr(image, boxes, txts, scores, font_path='/path/to/PaddleOCR/doc/fonts/simfang.ttf')
21
im_show = Image.fromarray(im_show)
22
im_show.save('result.jpg')

1
from paddleocr import PaddleOCR
2
#初始化模型
3
global_ocr = PaddleOCR(use_textline_orientation=True, #开启自动文本平放
4
                text_detection_model_name=text_detection_model_name,
5
                text_detection_model_dir=text_detection_model_dir,
6
                text_recognition_model_name=text_recognition_model_name,
7
                text_recognition_model_dir=text_recognition_model_dir,
8
                textline_orientation_model_name=textline_orientation_model_name,
9
                textline_orientation_model_dir=textline_orientation_model_dir,
10
                doc_orientation_classify_model_name=doc_orientation_classify_model_name,
11
                doc_orientation_classify_model_dir=doc_orientation_classify_model_dir,
12
                doc_unwarping_model_name=doc_unwarping_model_name,
13
                doc_unwarping_model_dir=doc_unwarping_model_dir,
14
                text_recognition_batch_size=50  # 识别批量处理数(CPU建议5-10,GPU可提至20)
15
                )
16
img_path = 'PaddleOCR/doc/imgs/11.jpg'
17
result = global_ocr.predict(img_path)
18
#自行处理返回的结果

1
def transpose_detector(img, cnt=0):
2
    '''
3
    检测人脸并裁剪,返回处理后的图像字节流(用于后续转Base64)
4
    :param img: imread后的图片对象
5
    :param cnt: 旋转次数计数器
6
    :return: 裁剪后的图像字节流(成功)/ None(失败)
7
    '''
8
    # 统一处理输入类型(包括上传文件)
9
    img = process_image_input(img)
10
    if img is None:
11
        return None
12

13
    # 初始化正脸检测器(移到函数内避免全局变量问题,也可保留全局)
14
    detector = dlib.get_frontal_face_detector()
15

16
    try:
17
        # 检测图上的人脸(原代码dector拼写错误,修正为detector)
18
        dets = detector(img, 1)
19
    except Exception as ex:
20
        print(f"人脸检测异常: {ex}")
21
        return None
22

23
    # 检测到人脸:裁剪并返回字节流
24
    if dets:
25
        face = dets[0]  # 获取第一个人脸(身份证默认单人脸)
26
        # 计算裁剪范围(原逻辑保留,增加边界检查避免越界)
27
        top = face.top() - 80 #40
28
        left = face.left() - 40 #15
29
        height = face.bottom() - face.top() + 120 #60
30
        width = face.right() - face.left() + 80 #40
31

32
        # 边界保护:避免裁剪范围超出原图尺寸(防止数组越界错误)
33
        top = max(0, top)
34
        left = max(0, left)
35
        bottom = min(img.shape[0], top + height)
36
        right = min(img.shape[1], left + width)
37
        # 重新计算有效裁剪尺寸(避免空白图)
38
        valid_height = bottom - top
39
        valid_width = right - left
40

41
        # 裁剪人脸区域(替换原双重循环,效率更高)
42
        cropped_face = img[top:bottom, left:right]
43

44
        # 检查裁剪结果是否有效(避免空图)
45
        if valid_height <= 0 or valid_width <= 0:
46
            print("裁剪范围无效,尝试旋转图像")
47
            return handle_rotate(img, cnt)
48

49
        # --------------- 关键:将裁剪图转为字节流(用于Base64)---------------
50
        # 1. 用cv2.imencode()将OpenCV图像转为PNG格式字节流(也可改为JPG,调整参数)
51
        # 参数1:格式后缀(.png/.jpg),参数2:图像数据,参数3:压缩质量(JPG可用,PNG忽略)
52
        success, img_byte_arr = cv2.imencode('.png', cropped_face)
53
        if not success:
54
            print("图像编码为字节流失败")
55
            return None
56

57
        # 2. 返回字节流(后续在主函数中转Base64)
58
        return img_byte_arr.tobytes()
59

60
    # 未检测到人脸:旋转图像重试(原逻辑保留)
61
    else:
62
        cnt += 1
63
        if cnt > 3:
64
            print("旋转超过3次仍未检测到人脸,处理失败")
65
            return None
66
        print(f"第{cnt}次旋转图像重试")
67

68
        # 图像旋转(原逻辑:transpose+flip 等效旋转90度,方向可根据需求调整)
69
        transpose_img = cv2.transpose(img)
70
        flipped_img = cv2.flip(transpose_img, 0)
71

72
        # 递归调用,继续检测旋转后的图像
73
        return transpose_detector(flipped_img, cnt)
74

75
def handle_rotate(img, cnt):
76
    '''辅助函数:处理裁剪无效时的图像旋转重试'''
77
    cnt += 1
78
    if cnt > 3:
79
        print("旋转超过3次仍未检测到有效人脸,处理失败")
80
        return None
81
    print(f"裁剪范围无效,第{cnt}次旋转图像重试")
82
    transpose_img = cv2.transpose(img)
83
    flipped_img = cv2.flip(transpose_img, 0)
84
    return transpose_detector(flipped_img, cnt)