文字识别OCR

图片的预处理，限制最大边的长度，并把宽和高变成32的倍数，方便之后的处理

import numpy


class DBProcessTest(object):
    """
    DB pre-process for Test mode
    """

    def __init__(self, params):
        # 最大边的长度，resize_type=0,调用resize_image_type0
        super(DBProcessTest, self).__init__()
        self.resize_type = 0
        if 'test_image_shape' in params:
            self.image_shape = params['test_image_shape']
            # print(self.image_shape)
            self.resize_type = 1
        if 'max_side_len' in params:
            self.max_side_len = params['max_side_len']
        else:
            self.max_side_len = 2400

    def resize_image_type0(self, im):
        """
        resize image to a size multiple of 32 which is required by the network
        args:
            img(array): array with shape [h, w, c]
        return(tuple):
            img, (ratio_h, ratio_w)
        """
        max_side_len = self.max_side_len
        h, w, _ = im.shape

        resize_w = w
        resize_h = h

        # limit the max side
        if max(resize_h, resize_w) > max_side_len:
            if resize_h > resize_w:
                ratio = float(max_side_len) / resize_h
            else:
                ratio = float(max_side_len) / resize_w
        else:
            ratio = 1.
        resize_h = int(resize_h * ratio)
        resize_w = int(resize_w * ratio)

        # resize image to a size multiple of 32
        if resize_h % 32 == 0:
            resize_h = resize_h
        elif resize_h // 32 <= 1:
            resize_h = 32
        else:
            resize_h = (resize_h // 32 - 1) * 32
        if resize_w % 32 == 0:
            resize_w = resize_w
        elif resize_w // 32 <= 1:
            resize_w = 32
        else:
            resize_w = (resize_w // 32 - 1) * 32
        try:
            if int(resize_w) <= 0 or int(resize_h) <= 0:
                return None, (None, None)
            im = cv2.resize(im, (int(resize_w), int(resize_h)))
        except:
            print(im.shape, resize_w, resize_h)
            sys.exit(0)
        ratio_h = resize_h / float(h)
        ratio_w = resize_w / float(w)
        return im, (ratio_h, ratio_w)

    def resize_image_type1(self, im):
        resize_h, resize_w = self.image_shape
        ori_h, ori_w = im.shape[:2]  # (h, w, c)
        im = cv2.resize(im, (int(resize_w), int(resize_h)))
        ratio_h = float(resize_h) / ori_h
        ratio_w = float(resize_w) / ori_w
        return im, (ratio_h, ratio_w)

    def normalize(self, im):
        img_mean = [0.485, 0.456, 0.406]
        img_std = [0.229, 0.224, 0.225]
        im = im.astype(np.float32, copy=False)
        im = im / 255
        im -= img_mean
        im /= img_std
        channel_swap = (2, 0, 1)
        im = im.transpose(channel_swap)
        return im

    def __call__(self, im):
        if self.resize_type == 0:
            im, (ratio_h, ratio_w) = self.resize_image_type0(im)  # resize
        else:
            im, (ratio_h, ratio_w) = self.resize_image_type1(im)
        im = self.normalize(im)
        im = im[np.newaxis, :]
        return [im, (ratio_h, ratio_w)]

参考：
https://github.com/PaddlePaddle/PaddleOCR/tree/5067126e8c84a104ca65a990a8521a5d59725161