Added code for PaddleOCR inference

src/models/thrid_party/paddleocr/infer/CTCLabelDecode.py

@@ -0,0 +1,215 @@
+import re
+import numpy as np
+import os
+from pathlib import Path
+
+
+class BaseRecLabelDecode(object):
+    """Convert between text-label and text-index"""
+
+    def __init__(self, character_dict_path=None, use_space_char=False):
+        cur_path = os.getcwd()
+        scriptDir = Path(__file__).resolve().parent
+        os.chdir(scriptDir)
+        character_dict_path = str(Path(scriptDir / "ppocr_keys_v1.txt"))
+
+        self.beg_str = "sos"
+        self.end_str = "eos"
+        self.reverse = False
+        self.character_str = []
+
+        if character_dict_path is None:
+            self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz"
+            dict_character = list(self.character_str)
+        else:
+            with open(character_dict_path, "rb") as fin:
+                lines = fin.readlines()
+                for line in lines:
+                    line = line.decode("utf-8").strip("\n").strip("\r\n")
+                    self.character_str.append(line)
+            if use_space_char:
+                self.character_str.append(" ")
+            dict_character = list(self.character_str)
+            if "arabic" in character_dict_path:
+                self.reverse = True
+
+        dict_character = self.add_special_char(dict_character)
+        self.dict = {}
+        for i, char in enumerate(dict_character):
+            self.dict[char] = i
+        self.character = dict_character
+        os.chdir(cur_path)
+
+    def pred_reverse(self, pred):
+        pred_re = []
+        c_current = ""
+        for c in pred:
+            if not bool(re.search("[a-zA-Z0-9 :*./%+-]", c)):
+                if c_current != "":
+                    pred_re.append(c_current)
+                pred_re.append(c)
+                c_current = ""
+            else:
+                c_current += c
+        if c_current != "":
+            pred_re.append(c_current)
+
+        return "".join(pred_re[::-1])
+
+    def add_special_char(self, dict_character):
+        return dict_character
+
+    def get_word_info(self, text, selection):
+        """
+        Group the decoded characters and record the corresponding decoded positions.
+
+        Args:
+            text: the decoded text
+            selection: the bool array that identifies which columns of features are decoded as non-separated characters
+        Returns:
+            word_list: list of the grouped words
+            word_col_list: list of decoding positions corresponding to each character in the grouped word
+            state_list: list of marker to identify the type of grouping words, including two types of grouping words:
+                        - 'cn': continous chinese characters (e.g., 你好啊)
+                        - 'en&num': continous english characters (e.g., hello), number (e.g., 123, 1.123), or mixed of them connected by '-' (e.g., VGG-16)
+                        The remaining characters in text are treated as separators between groups (e.g., space, '(', ')', etc.).
+        """
+        state = None
+        word_content = []
+        word_col_content = []
+        word_list = []
+        word_col_list = []
+        state_list = []
+        valid_col = np.where(selection == True)[0]
+
+        for c_i, char in enumerate(text):
+            if "\u4e00" <= char <= "\u9fff":
+                c_state = "cn"
+            elif bool(re.search("[a-zA-Z0-9]", char)):
+                c_state = "en&num"
+            else:
+                c_state = "splitter"
+
+            if (
+                char == "."
+                and state == "en&num"
+                and c_i + 1 < len(text)
+                and bool(re.search("[0-9]", text[c_i + 1]))
+            ):  # grouping floting number
+                c_state = "en&num"
+            if (
+                char == "-" and state == "en&num"
+            ):  # grouping word with '-', such as 'state-of-the-art'
+                c_state = "en&num"
+
+            if state == None:
+                state = c_state
+
+            if state != c_state:
+                if len(word_content) != 0:
+                    word_list.append(word_content)
+                    word_col_list.append(word_col_content)
+                    state_list.append(state)
+                    word_content = []
+                    word_col_content = []
+                state = c_state
+
+            if state != "splitter":
+                word_content.append(char)
+                word_col_content.append(valid_col[c_i])
+
+        if len(word_content) != 0:
+            word_list.append(word_content)
+            word_col_list.append(word_col_content)
+            state_list.append(state)
+
+        return word_list, word_col_list, state_list
+
+    def decode(
+        self,
+        text_index,
+        text_prob=None,
+        is_remove_duplicate=False,
+        return_word_box=False,
+    ):
+        """convert text-index into text-label."""
+        result_list = []
+        ignored_tokens = self.get_ignored_tokens()
+        batch_size = len(text_index)
+        for batch_idx in range(batch_size):
+            selection = np.ones(len(text_index[batch_idx]), dtype=bool)
+            if is_remove_duplicate:
+                selection[1:] = text_index[batch_idx][1:] != text_index[batch_idx][:-1]
+            for ignored_token in ignored_tokens:
+                selection &= text_index[batch_idx] != ignored_token
+
+            char_list = [
+                self.character[text_id] for text_id in text_index[batch_idx][selection]
+            ]
+            if text_prob is not None:
+                conf_list = text_prob[batch_idx][selection]
+            else:
+                conf_list = [1] * len(selection)
+            if len(conf_list) == 0:
+                conf_list = [0]
+
+            text = "".join(char_list)
+
+            if self.reverse:  # for arabic rec
+                text = self.pred_reverse(text)
+
+            if return_word_box:
+                word_list, word_col_list, state_list = self.get_word_info(
+                    text, selection
+                )
+                result_list.append(
+                    (
+                        text,
+                        np.mean(conf_list).tolist(),
+                        [
+                            len(text_index[batch_idx]),
+                            word_list,
+                            word_col_list,
+                            state_list,
+                        ],
+                    )
+                )
+            else:
+                result_list.append((text, np.mean(conf_list).tolist()))
+        return result_list
+
+    def get_ignored_tokens(self):
+        return [0]  # for ctc blank
+
+
+class CTCLabelDecode(BaseRecLabelDecode):
+    """Convert between text-label and text-index"""
+
+    def __init__(self, character_dict_path=None, use_space_char=False, **kwargs):
+        super(CTCLabelDecode, self).__init__(character_dict_path, use_space_char)
+
+    def __call__(self, preds, label=None, return_word_box=False, *args, **kwargs):
+        if isinstance(preds, tuple) or isinstance(preds, list):
+            preds = preds[-1]
+        assert isinstance(preds, np.ndarray)
+        preds_idx = preds.argmax(axis=2)
+        preds_prob = preds.max(axis=2)
+        text = self.decode(
+            preds_idx,
+            preds_prob,
+            is_remove_duplicate=True,
+            return_word_box=return_word_box,
+        )
+        if return_word_box:
+            for rec_idx, rec in enumerate(text):
+                wh_ratio = kwargs["wh_ratio_list"][rec_idx]
+                max_wh_ratio = kwargs["max_wh_ratio"]
+                rec[2][0] = rec[2][0] * (wh_ratio / max_wh_ratio)
+        if label is None:
+            return text
+        label = self.decode(label)
+        return text, label
+
+    def add_special_char(self, dict_character):
+        dict_character = ["blank"] + dict_character
+        return dict_character

src/models/thrid_party/paddleocr/infer/DBPostProcess.py

@@ -0,0 +1,229 @@
+import numpy as np
+import cv2
+
+from shapely.geometry import Polygon
+import pyclipper
+
+
+class DBPostProcess(object):
+    """
+    The post process for Differentiable Binarization (DB).
+    """
+
+    def __init__(
+        self,
+        thresh=0.3,
+        box_thresh=0.7,
+        max_candidates=1000,
+        unclip_ratio=2.0,
+        use_dilation=False,
+        score_mode="fast",
+        box_type="quad",
+        **kwargs
+    ):
+        self.thresh = thresh
+        self.box_thresh = box_thresh
+        self.max_candidates = max_candidates
+        self.unclip_ratio = unclip_ratio
+        self.min_size = 3
+        self.score_mode = score_mode
+        self.box_type = box_type
+        assert score_mode in [
+            "slow",
+            "fast",
+        ], "Score mode must be in [slow, fast] but got: {}".format(score_mode)
+
+        self.dilation_kernel = None if not use_dilation else np.array([[1, 1], [1, 1]])
+
+    def polygons_from_bitmap(self, pred, _bitmap, dest_width, dest_height):
+        """
+        _bitmap: single map with shape (1, H, W),
+            whose values are binarized as {0, 1}
+        """
+
+        bitmap = _bitmap
+        height, width = bitmap.shape
+
+        boxes = []
+        scores = []
+
+        contours, _ = cv2.findContours(
+            (bitmap * 255).astype(np.uint8), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE
+        )
+
+        for contour in contours[: self.max_candidates]:
+            epsilon = 0.002 * cv2.arcLength(contour, True)
+            approx = cv2.approxPolyDP(contour, epsilon, True)
+            points = approx.reshape((-1, 2))
+            if points.shape[0] < 4:
+                continue
+
+            score = self.box_score_fast(pred, points.reshape(-1, 2))
+            if self.box_thresh > score:
+                continue
+
+            if points.shape[0] > 2:
+                box = self.unclip(points, self.unclip_ratio)
+                if len(box) > 1:
+                    continue
+            else:
+                continue
+            box = box.reshape(-1, 2)
+
+            _, sside = self.get_mini_boxes(box.reshape((-1, 1, 2)))
+            if sside < self.min_size + 2:
+                continue
+
+            box = np.array(box)
+            box[:, 0] = np.clip(np.round(box[:, 0] / width * dest_width), 0, dest_width)
+            box[:, 1] = np.clip(
+                np.round(box[:, 1] / height * dest_height), 0, dest_height
+            )
+            boxes.append(box.tolist())
+            scores.append(score)
+        return boxes, scores
+
+    def boxes_from_bitmap(self, pred, _bitmap, dest_width, dest_height):
+        """
+        _bitmap: single map with shape (1, H, W),
+                whose values are binarized as {0, 1}
+        """
+
+        bitmap = _bitmap
+        height, width = bitmap.shape
+
+        outs = cv2.findContours(
+            (bitmap * 255).astype(np.uint8), cv2.RETR_LIST, cv2.CHAIN_APPROX_SIMPLE
+        )
+        if len(outs) == 3:
+            img, contours, _ = outs[0], outs[1], outs[2]
+        elif len(outs) == 2:
+            contours, _ = outs[0], outs[1]
+
+        num_contours = min(len(contours), self.max_candidates)
+
+        boxes = []
+        scores = []
+        for index in range(num_contours):
+            contour = contours[index]
+            points, sside = self.get_mini_boxes(contour)
+            if sside < self.min_size:
+                continue
+            points = np.array(points)
+            if self.score_mode == "fast":
+                score = self.box_score_fast(pred, points.reshape(-1, 2))
+            else:
+                score = self.box_score_slow(pred, contour)
+            if self.box_thresh > score:
+                continue
+
+            box = self.unclip(points, self.unclip_ratio).reshape(-1, 1, 2)
+            box, sside = self.get_mini_boxes(box)
+            if sside < self.min_size + 2:
+                continue
+            box = np.array(box)
+
+            box[:, 0] = np.clip(np.round(box[:, 0] / width * dest_width), 0, dest_width)
+            box[:, 1] = np.clip(
+                np.round(box[:, 1] / height * dest_height), 0, dest_height
+            )
+            boxes.append(box.astype("int32"))
+            scores.append(score)
+        return np.array(boxes, dtype="int32"), scores
+
+    def unclip(self, box, unclip_ratio):
+        poly = Polygon(box)
+        distance = poly.area * unclip_ratio / poly.length
+        offset = pyclipper.PyclipperOffset()
+        offset.AddPath(box, pyclipper.JT_ROUND, pyclipper.ET_CLOSEDPOLYGON)
+        expanded = np.array(offset.Execute(distance))
+        return expanded
+
+    def get_mini_boxes(self, contour):
+        bounding_box = cv2.minAreaRect(contour)
+        points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0])
+
+        index_1, index_2, index_3, index_4 = 0, 1, 2, 3
+        if points[1][1] > points[0][1]:
+            index_1 = 0
+            index_4 = 1
+        else:
+            index_1 = 1
+            index_4 = 0
+        if points[3][1] > points[2][1]:
+            index_2 = 2
+            index_3 = 3
+        else:
+            index_2 = 3
+            index_3 = 2
+
+        box = [points[index_1], points[index_2], points[index_3], points[index_4]]
+        return box, min(bounding_box[1])
+
+    def box_score_fast(self, bitmap, _box):
+        """
+        box_score_fast: use bbox mean score as the mean score
+        """
+        h, w = bitmap.shape[:2]
+        box = _box.copy()
+        xmin = np.clip(np.floor(box[:, 0].min()).astype("int32"), 0, w - 1)
+        xmax = np.clip(np.ceil(box[:, 0].max()).astype("int32"), 0, w - 1)
+        ymin = np.clip(np.floor(box[:, 1].min()).astype("int32"), 0, h - 1)
+        ymax = np.clip(np.ceil(box[:, 1].max()).astype("int32"), 0, h - 1)
+
+        mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8)
+        box[:, 0] = box[:, 0] - xmin
+        box[:, 1] = box[:, 1] - ymin
+        cv2.fillPoly(mask, box.reshape(1, -1, 2).astype("int32"), 1)
+        return cv2.mean(bitmap[ymin : ymax + 1, xmin : xmax + 1], mask)[0]
+
+    def box_score_slow(self, bitmap, contour):
+        """
+        box_score_slow: use polyon mean score as the mean score
+        """
+        h, w = bitmap.shape[:2]
+        contour = contour.copy()
+        contour = np.reshape(contour, (-1, 2))
+
+        xmin = np.clip(np.min(contour[:, 0]), 0, w - 1)
+        xmax = np.clip(np.max(contour[:, 0]), 0, w - 1)
+        ymin = np.clip(np.min(contour[:, 1]), 0, h - 1)
+        ymax = np.clip(np.max(contour[:, 1]), 0, h - 1)
+
+        mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8)
+
+        contour[:, 0] = contour[:, 0] - xmin
+        contour[:, 1] = contour[:, 1] - ymin
+
+        cv2.fillPoly(mask, contour.reshape(1, -1, 2).astype("int32"), 1)
+        return cv2.mean(bitmap[ymin : ymax + 1, xmin : xmax + 1], mask)[0]
+
+    def __call__(self, outs_dict, shape_list):
+        pred = outs_dict["maps"]
+        assert isinstance(pred, np.ndarray)
+        pred = pred[:, 0, :, :]
+        segmentation = pred > self.thresh
+
+        boxes_batch = []
+        for batch_index in range(pred.shape[0]):
+            src_h, src_w, ratio_h, ratio_w = shape_list[batch_index]
+            if self.dilation_kernel is not None:
+                mask = cv2.dilate(
+                    np.array(segmentation[batch_index]).astype(np.uint8),
+                    self.dilation_kernel,
+                )
+            else:
+                mask = segmentation[batch_index]
+            if self.box_type == "poly":
+                boxes, scores = self.polygons_from_bitmap(
+                    pred[batch_index], mask, src_w, src_h
+                )
+            elif self.box_type == "quad":
+                boxes, scores = self.boxes_from_bitmap(
+                    pred[batch_index], mask, src_w, src_h
+                )
+            else:
+                raise ValueError("box_type can only be one of ['quad', 'poly']")
+
+            boxes_batch.append({"points": boxes})
+        return boxes_batch

src/models/thrid_party/paddleocr/infer/operators.py

@@ -0,0 +1,186 @@
+import numpy as np
+import cv2
+import math
+import sys
+
+
+class DetResizeForTest(object):
+    def __init__(self, **kwargs):
+        super(DetResizeForTest, self).__init__()
+        self.resize_type = 0
+        self.keep_ratio = False
+        if "image_shape" in kwargs:
+            self.image_shape = kwargs["image_shape"]
+            self.resize_type = 1
+            if "keep_ratio" in kwargs:
+                self.keep_ratio = kwargs["keep_ratio"]
+        elif "limit_side_len" in kwargs:
+            self.limit_side_len = kwargs["limit_side_len"]
+            self.limit_type = kwargs.get("limit_type", "min")
+        elif "resize_long" in kwargs:
+            self.resize_type = 2
+            self.resize_long = kwargs.get("resize_long", 960)
+        else:
+            self.limit_side_len = 736
+            self.limit_type = "min"
+
+    def __call__(self, data):
+        img = data["image"]
+        src_h, src_w, _ = img.shape
+        if sum([src_h, src_w]) < 64:
+            img = self.image_padding(img)
+
+        if self.resize_type == 0:
+            # img, shape = self.resize_image_type0(img)
+            img, [ratio_h, ratio_w] = self.resize_image_type0(img)
+        elif self.resize_type == 2:
+            img, [ratio_h, ratio_w] = self.resize_image_type2(img)
+        else:
+            # img, shape = self.resize_image_type1(img)
+            img, [ratio_h, ratio_w] = self.resize_image_type1(img)
+        data["image"] = img
+        data["shape"] = np.array([src_h, src_w, ratio_h, ratio_w])
+        return data
+
+    def image_padding(self, im, value=0):
+        h, w, c = im.shape
+        im_pad = np.zeros((max(32, h), max(32, w), c), np.uint8) + value
+        im_pad[:h, :w, :] = im
+        return im_pad
+
+    def resize_image_type1(self, img):
+        resize_h, resize_w = self.image_shape
+        ori_h, ori_w = img.shape[:2]  # (h, w, c)
+        if self.keep_ratio is True:
+            resize_w = ori_w * resize_h / ori_h
+            N = math.ceil(resize_w / 32)
+            resize_w = N * 32
+        ratio_h = float(resize_h) / ori_h
+        ratio_w = float(resize_w) / ori_w
+        img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        # return img, np.array([ori_h, ori_w])
+        return img, [ratio_h, ratio_w]
+
+    def resize_image_type0(self, img):
+        """
+        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)
+        """
+        limit_side_len = self.limit_side_len
+        h, w, c = img.shape
+
+        # limit the max side
+        if self.limit_type == "max":
+            if max(h, w) > limit_side_len:
+                if h > w:
+                    ratio = float(limit_side_len) / h
+                else:
+                    ratio = float(limit_side_len) / w
+            else:
+                ratio = 1.0
+        elif self.limit_type == "min":
+            if min(h, w) < limit_side_len:
+                if h < w:
+                    ratio = float(limit_side_len) / h
+                else:
+                    ratio = float(limit_side_len) / w
+            else:
+                ratio = 1.0
+        elif self.limit_type == "resize_long":
+            ratio = float(limit_side_len) / max(h, w)
+        else:
+            raise Exception("not support limit type, image ")
+        resize_h = int(h * ratio)
+        resize_w = int(w * ratio)
+
+        resize_h = max(int(round(resize_h / 32) * 32), 32)
+        resize_w = max(int(round(resize_w / 32) * 32), 32)
+
+        try:
+            if int(resize_w) <= 0 or int(resize_h) <= 0:
+                return None, (None, None)
+            img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        except:
+            print(img.shape, resize_w, resize_h)
+            sys.exit(0)
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+        return img, [ratio_h, ratio_w]
+
+    def resize_image_type2(self, img):
+        h, w, _ = img.shape
+
+        resize_w = w
+        resize_h = h
+
+        if resize_h > resize_w:
+            ratio = float(self.resize_long) / resize_h
+        else:
+            ratio = float(self.resize_long) / resize_w
+
+        resize_h = int(resize_h * ratio)
+        resize_w = int(resize_w * ratio)
+
+        max_stride = 128
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        img = cv2.resize(img, (int(resize_w), int(resize_h)))
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+
+        return img, [ratio_h, ratio_w]
+
+
+class NormalizeImage(object):
+    """normalize image such as substract mean, divide std"""
+
+    def __init__(self, scale=None, mean=None, std=None, order="chw", **kwargs):
+        if isinstance(scale, str):
+            scale = eval(scale)
+        self.scale = np.float32(scale if scale is not None else 1.0 / 255.0)
+        mean = mean if mean is not None else [0.485, 0.456, 0.406]
+        std = std if std is not None else [0.229, 0.224, 0.225]
+
+        shape = (3, 1, 1) if order == "chw" else (1, 1, 3)
+        self.mean = np.array(mean).reshape(shape).astype("float32")
+        self.std = np.array(std).reshape(shape).astype("float32")
+
+    def __call__(self, data):
+        img = data["image"]
+        from PIL import Image
+
+        if isinstance(img, Image.Image):
+            img = np.array(img)
+        assert isinstance(img, np.ndarray), "invalid input 'img' in NormalizeImage"
+        data["image"] = (img.astype("float32") * self.scale - self.mean) / self.std
+        return data
+
+
+class ToCHWImage(object):
+    """convert hwc image to chw image"""
+
+    def __init__(self, **kwargs):
+        pass
+
+    def __call__(self, data):
+        img = data["image"]
+        from PIL import Image
+
+        if isinstance(img, Image.Image):
+            img = np.array(img)
+        data["image"] = img.transpose((2, 0, 1))
+        return data
+
+
+class KeepKeys(object):
+    def __init__(self, keep_keys, **kwargs):
+        self.keep_keys = keep_keys
+
+    def __call__(self, data):
+        data_list = []
+        for key in self.keep_keys:
+            data_list.append(data[key])
+        return data_list

src/models/thrid_party/paddleocr/infer/predict_det.py

@@ -0,0 +1,298 @@
+# 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
+
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.insert(0, os.path.abspath(os.path.join(__dir__, "../..")))
+
+os.environ["FLAGS_allocator_strategy"] = "auto_growth"
+
+import cv2
+import numpy as np
+import time
+import sys
+
+# import tools.infer.utility as utility
+import utility
+from utility import get_logger
+
+from DBPostProcess import DBPostProcess
+from operators import DetResizeForTest, KeepKeys, NormalizeImage, ToCHWImage
+
+
+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
+
+logger = get_logger()
+
+
+class TextDetector(object):
+    def __init__(self, args):
+        self.args = args
+        self.det_algorithm = args.det_algorithm
+        self.use_onnx = args.use_onnx
+        postprocess_params = {}
+        assert self.det_algorithm == "DB"
+        postprocess_params["name"] = "DBPostProcess"
+        postprocess_params["thresh"] = args.det_db_thresh
+        postprocess_params["box_thresh"] = args.det_db_box_thresh
+        postprocess_params["max_candidates"] = 1000
+        postprocess_params["unclip_ratio"] = args.det_db_unclip_ratio
+        postprocess_params["use_dilation"] = args.use_dilation
+        postprocess_params["score_mode"] = args.det_db_score_mode
+        postprocess_params["box_type"] = args.det_box_type
+
+        self.preprocess_op = [
+            DetResizeForTest(limit_side_len=args.det_limit_side_len, limit_type=args.det_limit_type),
+            NormalizeImage(std= [0.229, 0.224, 0.225], mean= [0.485, 0.456, 0.406], scale= 1./255., order= "hwc"),
+            ToCHWImage(),
+            KeepKeys(keep_keys= ["image", "shape"])
+        ]
+        self.postprocess_op = DBPostProcess(**postprocess_params)
+        (
+            self.predictor,
+            self.input_tensor,
+            self.output_tensors,
+            self.config,
+        ) = utility.create_predictor(args, "det", logger)
+
+        assert self.use_onnx
+        if self.use_onnx:
+            img_h, img_w = self.input_tensor.shape[2:]
+            if isinstance(img_h, str) or isinstance(img_w, str):
+                pass
+            elif img_h is not None and img_w is not None and img_h > 0 and img_w > 0:
+                self.preprocess_op[0] = DetResizeForTest(image_shape=[img_h, img_w])
+
+
+    def order_points_clockwise(self, pts):
+        rect = np.zeros((4, 2), dtype="float32")
+        s = pts.sum(axis=1)
+        rect[0] = pts[np.argmin(s)]
+        rect[2] = pts[np.argmax(s)]
+        tmp = np.delete(pts, (np.argmin(s), np.argmax(s)), axis=0)
+        diff = np.diff(np.array(tmp), axis=1)
+        rect[1] = tmp[np.argmin(diff)]
+        rect[3] = tmp[np.argmax(diff)]
+        return rect
+
+    def clip_det_res(self, points, img_height, img_width):
+        for pno in range(points.shape[0]):
+            points[pno, 0] = int(min(max(points[pno, 0], 0), img_width - 1))
+            points[pno, 1] = int(min(max(points[pno, 1], 0), img_height - 1))
+        return points
+
+    def filter_tag_det_res(self, dt_boxes, image_shape):
+        img_height, img_width = image_shape[0:2]
+        dt_boxes_new = []
+        for box in dt_boxes:
+            if type(box) is list:
+                box = np.array(box)
+            box = self.order_points_clockwise(box)
+            box = self.clip_det_res(box, img_height, img_width)
+            rect_width = int(np.linalg.norm(box[0] - box[1]))
+            rect_height = int(np.linalg.norm(box[0] - box[3]))
+            if rect_width <= 3 or rect_height <= 3:
+                continue
+            dt_boxes_new.append(box)
+        dt_boxes = np.array(dt_boxes_new)
+        return dt_boxes
+
+    def filter_tag_det_res_only_clip(self, dt_boxes, image_shape):
+        img_height, img_width = image_shape[0:2]
+        dt_boxes_new = []
+        for box in dt_boxes:
+            if type(box) is list:
+                box = np.array(box)
+            box = self.clip_det_res(box, img_height, img_width)
+            dt_boxes_new.append(box)
+        dt_boxes = np.array(dt_boxes_new)
+        return dt_boxes
+
+    def predict(self, img):
+        ori_im = img.copy()
+        data = {"image": img}
+
+        st = time.time()
+
+        if self.args.benchmark:
+            self.autolog.times.start()
+
+        data = transform(data, self.preprocess_op)
+        img, shape_list = data
+        if img is None:
+            return None, 0
+        img = np.expand_dims(img, axis=0)
+        shape_list = np.expand_dims(shape_list, axis=0)
+        img = img.copy()
+
+        if self.args.benchmark:
+            self.autolog.times.stamp()
+        if self.use_onnx:
+            input_dict = {}
+            input_dict[self.input_tensor.name] = img
+            outputs = self.predictor.run(self.output_tensors, input_dict)
+        else:
+            self.input_tensor.copy_from_cpu(img)
+            self.predictor.run()
+            outputs = []
+            for output_tensor in self.output_tensors:
+                output = output_tensor.copy_to_cpu()
+                outputs.append(output)
+            if self.args.benchmark:
+                self.autolog.times.stamp()
+
+        preds = {}
+        if self.det_algorithm == "EAST":
+            preds["f_geo"] = outputs[0]
+            preds["f_score"] = outputs[1]
+        elif self.det_algorithm == "SAST":
+            preds["f_border"] = outputs[0]
+            preds["f_score"] = outputs[1]
+            preds["f_tco"] = outputs[2]
+            preds["f_tvo"] = outputs[3]
+        elif self.det_algorithm in ["DB", "PSE", "DB++"]:
+            preds["maps"] = outputs[0]
+        elif self.det_algorithm == "FCE":
+            for i, output in enumerate(outputs):
+                preds["level_{}".format(i)] = output
+        elif self.det_algorithm == "CT":
+            preds["maps"] = outputs[0]
+            preds["score"] = outputs[1]
+        else:
+            raise NotImplementedError
+
+        post_result = self.postprocess_op(preds, shape_list)
+        dt_boxes = post_result[0]["points"]
+
+        if self.args.det_box_type == "poly":
+            dt_boxes = self.filter_tag_det_res_only_clip(dt_boxes, ori_im.shape)
+        else:
+            dt_boxes = self.filter_tag_det_res(dt_boxes, ori_im.shape)
+
+        if self.args.benchmark:
+            self.autolog.times.end(stamp=True)
+        et = time.time()
+        return dt_boxes, et - st
+
+    def __call__(self, img):
+        # For image like poster with one side much greater than the other side,
+        # splitting recursively and processing with overlap to enhance performance.
+        MIN_BOUND_DISTANCE = 50
+        dt_boxes = np.zeros((0, 4, 2), dtype=np.float32)
+        elapse = 0
+        if (
+            img.shape[0] / img.shape[1] > 2
+            and img.shape[0] > self.args.det_limit_side_len
+        ):
+            start_h = 0
+            end_h = 0
+            while end_h <= img.shape[0]:
+                end_h = start_h + img.shape[1] * 3 // 4
+                subimg = img[start_h:end_h, :]
+                if len(subimg) == 0:
+                    break
+                sub_dt_boxes, sub_elapse = self.predict(subimg)
+                offset = start_h
+                # To prevent text blocks from being cut off, roll back a certain buffer area.
+                if (
+                    len(sub_dt_boxes) == 0
+                    or img.shape[1] - max([x[-1][1] for x in sub_dt_boxes])
+                    > MIN_BOUND_DISTANCE
+                ):
+                    start_h = end_h
+                else:
+                    sorted_indices = np.argsort(sub_dt_boxes[:, 2, 1])
+                    sub_dt_boxes = sub_dt_boxes[sorted_indices]
+                    bottom_line = (
+                        0
+                        if len(sub_dt_boxes) <= 1
+                        else int(np.max(sub_dt_boxes[:-1, 2, 1]))
+                    )
+                    if bottom_line > 0:
+                        start_h += bottom_line
+                        sub_dt_boxes = sub_dt_boxes[
+                            sub_dt_boxes[:, 2, 1] <= bottom_line
+                        ]
+                    else:
+                        start_h = end_h
+                if len(sub_dt_boxes) > 0:
+                    if dt_boxes.shape[0] == 0:
+                        dt_boxes = sub_dt_boxes + np.array(
+                            [0, offset], dtype=np.float32
+                        )
+                    else:
+                        dt_boxes = np.append(
+                            dt_boxes,
+                            sub_dt_boxes + np.array([0, offset], dtype=np.float32),
+                            axis=0,
+                        )
+                elapse += sub_elapse
+        elif (
+            img.shape[1] / img.shape[0] > 3
+            and img.shape[1] > self.args.det_limit_side_len * 3
+        ):
+            start_w = 0
+            end_w = 0
+            while end_w <= img.shape[1]:
+                end_w = start_w + img.shape[0] * 3 // 4
+                subimg = img[:, start_w:end_w]
+                if len(subimg) == 0:
+                    break
+                sub_dt_boxes, sub_elapse = self.predict(subimg)
+                offset = start_w
+                if (
+                    len(sub_dt_boxes) == 0
+                    or img.shape[0] - max([x[-1][0] for x in sub_dt_boxes])
+                    > MIN_BOUND_DISTANCE
+                ):
+                    start_w = end_w
+                else:
+                    sorted_indices = np.argsort(sub_dt_boxes[:, 2, 0])
+                    sub_dt_boxes = sub_dt_boxes[sorted_indices]
+                    right_line = (
+                        0
+                        if len(sub_dt_boxes) <= 1
+                        else int(np.max(sub_dt_boxes[:-1, 1, 0]))
+                    )
+                    if right_line > 0:
+                        start_w += right_line
+                        sub_dt_boxes = sub_dt_boxes[sub_dt_boxes[:, 1, 0] <= right_line]
+                    else:
+                        start_w = end_w
+                if len(sub_dt_boxes) > 0:
+                    if dt_boxes.shape[0] == 0:
+                        dt_boxes = sub_dt_boxes + np.array(
+                            [offset, 0], dtype=np.float32
+                        )
+                    else:
+                        dt_boxes = np.append(
+                            dt_boxes,
+                            sub_dt_boxes + np.array([offset, 0], dtype=np.float32),
+                            axis=0,
+                        )
+                elapse += sub_elapse
+        else:
+            dt_boxes, elapse = self.predict(img)
+        return dt_boxes, elapse
+

src/models/thrid_party/paddleocr/infer/predict_rec.py

@@ -0,0 +1,383 @@
+# 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
+from PIL import Image
+
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.insert(0, os.path.abspath(os.path.join(__dir__, "../..")))
+
+os.environ["FLAGS_allocator_strategy"] = "auto_growth"
+
+import cv2
+import numpy as np
+import math
+import time
+
+import utility
+from utility import get_logger
+
+from CTCLabelDecode import CTCLabelDecode
+
+logger = get_logger()
+
+
+class TextRecognizer(object):
+    def __init__(self, args):
+        self.rec_image_shape = [int(v) for v in args.rec_image_shape.split(",")]
+        self.rec_batch_num = args.rec_batch_num
+        self.rec_algorithm = args.rec_algorithm
+        self.postprocess_op = CTCLabelDecode(character_dict_path=args.rec_char_dict_path, use_space_char=args.use_space_char)
+        (
+            self.predictor,
+            self.input_tensor,
+            self.output_tensors,
+            self.config,
+        ) = utility.create_predictor(args, "rec", logger)
+        self.benchmark = args.benchmark
+        self.use_onnx = args.use_onnx
+        self.return_word_box = args.return_word_box
+
+    def resize_norm_img(self, img, max_wh_ratio):
+        imgC, imgH, imgW = self.rec_image_shape
+        if self.rec_algorithm == "NRTR" or self.rec_algorithm == "ViTSTR":
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+            # return padding_im
+            image_pil = Image.fromarray(np.uint8(img))
+            if self.rec_algorithm == "ViTSTR":
+                img = image_pil.resize([imgW, imgH], Image.BICUBIC)
+            else:
+                img = image_pil.resize([imgW, imgH], Image.Resampling.LANCZOS)
+            img = np.array(img)
+            norm_img = np.expand_dims(img, -1)
+            norm_img = norm_img.transpose((2, 0, 1))
+            if self.rec_algorithm == "ViTSTR":
+                norm_img = norm_img.astype(np.float32) / 255.0
+            else:
+                norm_img = norm_img.astype(np.float32) / 128.0 - 1.0
+            return norm_img
+        elif self.rec_algorithm == "RFL":
+            img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+            resized_image = cv2.resize(img, (imgW, imgH), interpolation=cv2.INTER_CUBIC)
+            resized_image = resized_image.astype("float32")
+            resized_image = resized_image / 255
+            resized_image = resized_image[np.newaxis, :]
+            resized_image -= 0.5
+            resized_image /= 0.5
+            return resized_image
+
+        assert imgC == img.shape[2]
+        imgW = int((imgH * max_wh_ratio))
+        if self.use_onnx:
+            w = self.input_tensor.shape[3:][0]
+            if isinstance(w, str):
+                pass
+            elif w is not None and w > 0:
+                imgW = w
+        h, w = img.shape[:2]
+        ratio = w / float(h)
+        if math.ceil(imgH * ratio) > imgW:
+            resized_w = imgW
+        else:
+            resized_w = int(math.ceil(imgH * ratio))
+        if self.rec_algorithm == "RARE":
+            if resized_w > self.rec_image_shape[2]:
+                resized_w = self.rec_image_shape[2]
+            imgW = self.rec_image_shape[2]
+        resized_image = cv2.resize(img, (resized_w, imgH))
+        resized_image = resized_image.astype("float32")
+        resized_image = resized_image.transpose((2, 0, 1)) / 255
+        resized_image -= 0.5
+        resized_image /= 0.5
+        padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32)
+        padding_im[:, :, 0:resized_w] = resized_image
+        return padding_im
+
+    def resize_norm_img_vl(self, img, image_shape):
+        imgC, imgH, imgW = image_shape
+        img = img[:, :, ::-1]  # bgr2rgb
+        resized_image = cv2.resize(img, (imgW, imgH), interpolation=cv2.INTER_LINEAR)
+        resized_image = resized_image.astype("float32")
+        resized_image = resized_image.transpose((2, 0, 1)) / 255
+        return resized_image
+
+    def resize_norm_img_srn(self, img, image_shape):
+        imgC, imgH, imgW = image_shape
+
+        img_black = np.zeros((imgH, imgW))
+        im_hei = img.shape[0]
+        im_wid = img.shape[1]
+
+        if im_wid <= im_hei * 1:
+            img_new = cv2.resize(img, (imgH * 1, imgH))
+        elif im_wid <= im_hei * 2:
+            img_new = cv2.resize(img, (imgH * 2, imgH))
+        elif im_wid <= im_hei * 3:
+            img_new = cv2.resize(img, (imgH * 3, imgH))
+        else:
+            img_new = cv2.resize(img, (imgW, imgH))
+
+        img_np = np.asarray(img_new)
+        img_np = cv2.cvtColor(img_np, cv2.COLOR_BGR2GRAY)
+        img_black[:, 0 : img_np.shape[1]] = img_np
+        img_black = img_black[:, :, np.newaxis]
+
+        row, col, c = img_black.shape
+        c = 1
+
+        return np.reshape(img_black, (c, row, col)).astype(np.float32)
+
+    def srn_other_inputs(self, image_shape, num_heads, max_text_length):
+        imgC, imgH, imgW = image_shape
+        feature_dim = int((imgH / 8) * (imgW / 8))
+
+        encoder_word_pos = (
+            np.array(range(0, feature_dim)).reshape((feature_dim, 1)).astype("int64")
+        )
+        gsrm_word_pos = (
+            np.array(range(0, max_text_length))
+            .reshape((max_text_length, 1))
+            .astype("int64")
+        )
+
+        gsrm_attn_bias_data = np.ones((1, max_text_length, max_text_length))
+        gsrm_slf_attn_bias1 = np.triu(gsrm_attn_bias_data, 1).reshape(
+            [-1, 1, max_text_length, max_text_length]
+        )
+        gsrm_slf_attn_bias1 = np.tile(gsrm_slf_attn_bias1, [1, num_heads, 1, 1]).astype(
+            "float32"
+        ) * [-1e9]
+
+        gsrm_slf_attn_bias2 = np.tril(gsrm_attn_bias_data, -1).reshape(
+            [-1, 1, max_text_length, max_text_length]
+        )
+        gsrm_slf_attn_bias2 = np.tile(gsrm_slf_attn_bias2, [1, num_heads, 1, 1]).astype(
+            "float32"
+        ) * [-1e9]
+
+        encoder_word_pos = encoder_word_pos[np.newaxis, :]
+        gsrm_word_pos = gsrm_word_pos[np.newaxis, :]
+
+        return [
+            encoder_word_pos,
+            gsrm_word_pos,
+            gsrm_slf_attn_bias1,
+            gsrm_slf_attn_bias2,
+        ]
+
+    def process_image_srn(self, img, image_shape, num_heads, max_text_length):
+        norm_img = self.resize_norm_img_srn(img, image_shape)
+        norm_img = norm_img[np.newaxis, :]
+
+        [
+            encoder_word_pos,
+            gsrm_word_pos,
+            gsrm_slf_attn_bias1,
+            gsrm_slf_attn_bias2,
+        ] = self.srn_other_inputs(image_shape, num_heads, max_text_length)
+
+        gsrm_slf_attn_bias1 = gsrm_slf_attn_bias1.astype(np.float32)
+        gsrm_slf_attn_bias2 = gsrm_slf_attn_bias2.astype(np.float32)
+        encoder_word_pos = encoder_word_pos.astype(np.int64)
+        gsrm_word_pos = gsrm_word_pos.astype(np.int64)
+
+        return (
+            norm_img,
+            encoder_word_pos,
+            gsrm_word_pos,
+            gsrm_slf_attn_bias1,
+            gsrm_slf_attn_bias2,
+        )
+
+    def resize_norm_img_sar(self, img, image_shape, width_downsample_ratio=0.25):
+        imgC, imgH, imgW_min, imgW_max = image_shape
+        h = img.shape[0]
+        w = img.shape[1]
+        valid_ratio = 1.0
+        # make sure new_width is an integral multiple of width_divisor.
+        width_divisor = int(1 / width_downsample_ratio)
+        # resize
+        ratio = w / float(h)
+        resize_w = math.ceil(imgH * ratio)
+        if resize_w % width_divisor != 0:
+            resize_w = round(resize_w / width_divisor) * width_divisor
+        if imgW_min is not None:
+            resize_w = max(imgW_min, resize_w)
+        if imgW_max is not None:
+            valid_ratio = min(1.0, 1.0 * resize_w / imgW_max)
+            resize_w = min(imgW_max, resize_w)
+        resized_image = cv2.resize(img, (resize_w, imgH))
+        resized_image = resized_image.astype("float32")
+        # norm
+        if image_shape[0] == 1:
+            resized_image = resized_image / 255
+            resized_image = resized_image[np.newaxis, :]
+        else:
+            resized_image = resized_image.transpose((2, 0, 1)) / 255
+        resized_image -= 0.5
+        resized_image /= 0.5
+        resize_shape = resized_image.shape
+        padding_im = -1.0 * np.ones((imgC, imgH, imgW_max), dtype=np.float32)
+        padding_im[:, :, 0:resize_w] = resized_image
+        pad_shape = padding_im.shape
+
+        return padding_im, resize_shape, pad_shape, valid_ratio
+
+    def resize_norm_img_spin(self, img):
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
+        # return padding_im
+        img = cv2.resize(img, tuple([100, 32]), cv2.INTER_CUBIC)
+        img = np.array(img, np.float32)
+        img = np.expand_dims(img, -1)
+        img = img.transpose((2, 0, 1))
+        mean = [127.5]
+        std = [127.5]
+        mean = np.array(mean, dtype=np.float32)
+        std = np.array(std, dtype=np.float32)
+        mean = np.float32(mean.reshape(1, -1))
+        stdinv = 1 / np.float32(std.reshape(1, -1))
+        img -= mean
+        img *= stdinv
+        return img
+
+    def resize_norm_img_svtr(self, img, image_shape):
+        imgC, imgH, imgW = image_shape
+        resized_image = cv2.resize(img, (imgW, imgH), interpolation=cv2.INTER_LINEAR)
+        resized_image = resized_image.astype("float32")
+        resized_image = resized_image.transpose((2, 0, 1)) / 255
+        resized_image -= 0.5
+        resized_image /= 0.5
+        return resized_image
+
+    def resize_norm_img_cppd_padding(
+        self, img, image_shape, padding=True, interpolation=cv2.INTER_LINEAR
+    ):
+        imgC, imgH, imgW = image_shape
+        h = img.shape[0]
+        w = img.shape[1]
+        if not padding:
+            resized_image = cv2.resize(img, (imgW, imgH), interpolation=interpolation)
+            resized_w = imgW
+        else:
+            ratio = w / float(h)
+            if math.ceil(imgH * ratio) > imgW:
+                resized_w = imgW
+            else:
+                resized_w = int(math.ceil(imgH * ratio))
+            resized_image = cv2.resize(img, (resized_w, imgH))
+        resized_image = resized_image.astype("float32")
+        if image_shape[0] == 1:
+            resized_image = resized_image / 255
+            resized_image = resized_image[np.newaxis, :]
+        else:
+            resized_image = resized_image.transpose((2, 0, 1)) / 255
+        resized_image -= 0.5
+        resized_image /= 0.5
+        padding_im = np.zeros((imgC, imgH, imgW), dtype=np.float32)
+        padding_im[:, :, 0:resized_w] = resized_image
+
+        return padding_im
+
+    def resize_norm_img_abinet(self, img, image_shape):
+        imgC, imgH, imgW = image_shape
+
+        resized_image = cv2.resize(img, (imgW, imgH), interpolation=cv2.INTER_LINEAR)
+        resized_image = resized_image.astype("float32")
+        resized_image = resized_image / 255.0
+
+        mean = np.array([0.485, 0.456, 0.406])
+        std = np.array([0.229, 0.224, 0.225])
+        resized_image = (resized_image - mean[None, None, ...]) / std[None, None, ...]
+        resized_image = resized_image.transpose((2, 0, 1))
+        resized_image = resized_image.astype("float32")
+
+        return resized_image
+
+    def norm_img_can(self, img, image_shape):
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)  # CAN only predict gray scale image
+
+        if self.inverse:
+            img = 255 - img
+
+        if self.rec_image_shape[0] == 1:
+            h, w = img.shape
+            _, imgH, imgW = self.rec_image_shape
+            if h < imgH or w < imgW:
+                padding_h = max(imgH - h, 0)
+                padding_w = max(imgW - w, 0)
+                img_padded = np.pad(
+                    img,
+                    ((0, padding_h), (0, padding_w)),
+                    "constant",
+                    constant_values=(255),
+                )
+                img = img_padded
+
+        img = np.expand_dims(img, 0) / 255.0  # h,w,c -> c,h,w
+        img = img.astype("float32")
+
+        return img
+
+    def __call__(self, img_list):
+        img_num = len(img_list)
+        # Calculate the aspect ratio of all text bars
+        width_list = []
+        for img in img_list:
+            width_list.append(img.shape[1] / float(img.shape[0]))
+        # Sorting can speed up the recognition process
+        indices = np.argsort(np.array(width_list))
+        rec_res = [["", 0.0]] * img_num
+        batch_num = self.rec_batch_num
+        st = time.time()
+        if self.benchmark:
+            self.autolog.times.start()
+        for beg_img_no in range(0, img_num, batch_num):
+            end_img_no = min(img_num, beg_img_no + batch_num)
+            norm_img_batch = []
+            imgC, imgH, imgW = self.rec_image_shape[:3]
+            max_wh_ratio = imgW / imgH
+            wh_ratio_list = []
+            for ino in range(beg_img_no, end_img_no):
+                h, w = img_list[indices[ino]].shape[0:2]
+                wh_ratio = w * 1.0 / h
+                max_wh_ratio = max(max_wh_ratio, wh_ratio)
+                wh_ratio_list.append(wh_ratio)
+            for ino in range(beg_img_no, end_img_no):
+                norm_img = self.resize_norm_img(
+                    img_list[indices[ino]], max_wh_ratio
+                )
+                norm_img = norm_img[np.newaxis, :]
+                norm_img_batch.append(norm_img)
+            norm_img_batch = np.concatenate(norm_img_batch)
+            norm_img_batch = norm_img_batch.copy()
+            if self.benchmark:
+                self.autolog.times.stamp()
+
+            assert self.use_onnx
+            input_dict = {}
+            input_dict[self.input_tensor.name] = norm_img_batch
+            outputs = self.predictor.run(self.output_tensors, input_dict)
+            preds = outputs[0]
+            rec_result = self.postprocess_op(
+                preds,
+                return_word_box=self.return_word_box,
+                wh_ratio_list=wh_ratio_list,
+                max_wh_ratio=max_wh_ratio,
+            )
+            for rno in range(len(rec_result)):
+                rec_res[indices[beg_img_no + rno]] = rec_result[rno]
+            if self.benchmark:
+                self.autolog.times.end(stamp=True)
+        return rec_res, time.time() - st

src/models/thrid_party/paddleocr/infer/utility.py

@@ -0,0 +1,713 @@
+# 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 argparse
+import os
+import sys
+import functools
+import logging
+import cv2
+import numpy as np
+import PIL
+from PIL import Image, ImageDraw, ImageFont
+import math
+import random
+
+
+def str2bool(v):
+    return v.lower() in ("true", "yes", "t", "y", "1")
+
+
+def str2int_tuple(v):
+    return tuple([int(i.strip()) for i in v.split(",")])
+
+
+def init_args():
+    parser = argparse.ArgumentParser()
+    # params for prediction engine
+    parser.add_argument("--use_gpu", type=str2bool, default=True)
+    parser.add_argument("--use_xpu", type=str2bool, default=False)
+    parser.add_argument("--use_npu", type=str2bool, default=False)
+    parser.add_argument("--use_mlu", type=str2bool, default=False)
+    parser.add_argument("--ir_optim", type=str2bool, default=True)
+    parser.add_argument("--use_tensorrt", type=str2bool, default=False)
+    parser.add_argument("--min_subgraph_size", type=int, default=15)
+    parser.add_argument("--precision", type=str, default="fp32")
+    parser.add_argument("--gpu_mem", type=int, default=500)
+    parser.add_argument("--gpu_id", type=int, default=0)
+
+    # params for text detector
+    parser.add_argument("--image_dir", type=str)
+    parser.add_argument("--page_num", type=int, default=0)
+    parser.add_argument("--det_algorithm", type=str, default="DB")
+    parser.add_argument("--det_model_dir", type=str)
+    parser.add_argument("--det_limit_side_len", type=float, default=960)
+    parser.add_argument("--det_limit_type", type=str, default="max")
+    parser.add_argument("--det_box_type", type=str, default="quad")
+
+    # DB parmas
+    parser.add_argument("--det_db_thresh", type=float, default=0.3)
+    parser.add_argument("--det_db_box_thresh", type=float, default=0.6)
+    parser.add_argument("--det_db_unclip_ratio", type=float, default=1.5)
+    parser.add_argument("--max_batch_size", type=int, default=10)
+    parser.add_argument("--use_dilation", type=str2bool, default=False)
+    parser.add_argument("--det_db_score_mode", type=str, default="fast")
+
+    # EAST parmas
+    parser.add_argument("--det_east_score_thresh", type=float, default=0.8)
+    parser.add_argument("--det_east_cover_thresh", type=float, default=0.1)
+    parser.add_argument("--det_east_nms_thresh", type=float, default=0.2)
+
+    # SAST parmas
+    parser.add_argument("--det_sast_score_thresh", type=float, default=0.5)
+    parser.add_argument("--det_sast_nms_thresh", type=float, default=0.2)
+
+    # PSE parmas
+    parser.add_argument("--det_pse_thresh", type=float, default=0)
+    parser.add_argument("--det_pse_box_thresh", type=float, default=0.85)
+    parser.add_argument("--det_pse_min_area", type=float, default=16)
+    parser.add_argument("--det_pse_scale", type=int, default=1)
+
+    # FCE parmas
+    parser.add_argument("--scales", type=list, default=[8, 16, 32])
+    parser.add_argument("--alpha", type=float, default=1.0)
+    parser.add_argument("--beta", type=float, default=1.0)
+    parser.add_argument("--fourier_degree", type=int, default=5)
+
+    # params for text recognizer
+    parser.add_argument("--rec_algorithm", type=str, default="SVTR_LCNet")
+    parser.add_argument("--rec_model_dir", type=str)
+    parser.add_argument("--rec_image_inverse", type=str2bool, default=True)
+    parser.add_argument("--rec_image_shape", type=str, default="3, 48, 320")
+    parser.add_argument("--rec_batch_num", type=int, default=6)
+    parser.add_argument("--max_text_length", type=int, default=25)
+    parser.add_argument(
+        "--rec_char_dict_path", type=str, default="./ppocr_keys_v1.txt"
+    )
+    parser.add_argument("--use_space_char", type=str2bool, default=True)
+    parser.add_argument("--vis_font_path", type=str, default="./doc/fonts/simfang.ttf")
+    parser.add_argument("--drop_score", type=float, default=0.5)
+
+    # params for e2e
+    parser.add_argument("--e2e_algorithm", type=str, default="PGNet")
+    parser.add_argument("--e2e_model_dir", type=str)
+    parser.add_argument("--e2e_limit_side_len", type=float, default=768)
+    parser.add_argument("--e2e_limit_type", type=str, default="max")
+
+    # PGNet parmas
+    parser.add_argument("--e2e_pgnet_score_thresh", type=float, default=0.5)
+    parser.add_argument(
+        "--e2e_char_dict_path", type=str, default="./ppocr/utils/ic15_dict.txt"
+    )
+    parser.add_argument("--e2e_pgnet_valid_set", type=str, default="totaltext")
+    parser.add_argument("--e2e_pgnet_mode", type=str, default="fast")
+
+    # params for text classifier
+    parser.add_argument("--use_angle_cls", type=str2bool, default=False)
+    parser.add_argument("--cls_model_dir", type=str)
+    parser.add_argument("--cls_image_shape", type=str, default="3, 48, 192")
+    parser.add_argument("--label_list", type=list, default=["0", "180"])
+    parser.add_argument("--cls_batch_num", type=int, default=6)
+    parser.add_argument("--cls_thresh", type=float, default=0.9)
+
+    parser.add_argument("--enable_mkldnn", type=str2bool, default=False)
+    parser.add_argument("--cpu_threads", type=int, default=10)
+    parser.add_argument("--use_pdserving", type=str2bool, default=False)
+    parser.add_argument("--warmup", type=str2bool, default=False)
+
+    # SR parmas
+    parser.add_argument("--sr_model_dir", type=str)
+    parser.add_argument("--sr_image_shape", type=str, default="3, 32, 128")
+    parser.add_argument("--sr_batch_num", type=int, default=1)
+
+    #
+    parser.add_argument("--draw_img_save_dir", type=str, default="./inference_results")
+    parser.add_argument("--save_crop_res", type=str2bool, default=False)
+    parser.add_argument("--crop_res_save_dir", type=str, default="./output")
+
+    # multi-process
+    parser.add_argument("--use_mp", type=str2bool, default=False)
+    parser.add_argument("--total_process_num", type=int, default=1)
+    parser.add_argument("--process_id", type=int, default=0)
+
+    parser.add_argument("--benchmark", type=str2bool, default=False)
+    parser.add_argument("--save_log_path", type=str, default="./log_output/")
+
+    parser.add_argument("--show_log", type=str2bool, default=True)
+    parser.add_argument("--use_onnx", type=str2bool, default=False)
+
+    # extended function
+    parser.add_argument(
+        "--return_word_box",
+        type=str2bool,
+        default=False,
+        help="Whether return the bbox of each word (split by space) or chinese character. Only used in ppstructure for layout recovery",
+    )
+
+    return parser
+
+
+def parse_args():
+    parser = init_args()
+    return parser.parse_args([])
+
+
+def create_predictor(args, mode, logger):
+    if mode == "det":
+        model_dir = args.det_model_dir
+    elif mode == "cls":
+        model_dir = args.cls_model_dir
+    elif mode == "rec":
+        model_dir = args.rec_model_dir
+    elif mode == "table":
+        model_dir = args.table_model_dir
+    elif mode == "ser":
+        model_dir = args.ser_model_dir
+    elif mode == "re":
+        model_dir = args.re_model_dir
+    elif mode == "sr":
+        model_dir = args.sr_model_dir
+    elif mode == "layout":
+        model_dir = args.layout_model_dir
+    else:
+        model_dir = args.e2e_model_dir
+
+    if model_dir is None:
+        logger.info("not find {} model file path {}".format(mode, model_dir))
+        sys.exit(0)
+    assert args.use_onnx
+
+    import onnxruntime as ort
+
+    model_file_path = model_dir
+    if not os.path.exists(model_file_path):
+        raise ValueError("not find model file path {}".format(model_file_path))
+    if args.use_gpu:
+        sess = ort.InferenceSession(
+            model_file_path, providers=["CUDAExecutionProvider"]
+        )
+    else:
+        sess = ort.InferenceSession(model_file_path)
+    return sess, sess.get_inputs()[0], None, None
+
+
+
+def get_output_tensors(args, mode, predictor):
+    output_names = predictor.get_output_names()
+    output_tensors = []
+    if mode == "rec" and args.rec_algorithm in ["CRNN", "SVTR_LCNet", "SVTR_HGNet"]:
+        output_name = "softmax_0.tmp_0"
+        if output_name in output_names:
+            return [predictor.get_output_handle(output_name)]
+        else:
+            for output_name in output_names:
+                output_tensor = predictor.get_output_handle(output_name)
+                output_tensors.append(output_tensor)
+    else:
+        for output_name in output_names:
+            output_tensor = predictor.get_output_handle(output_name)
+            output_tensors.append(output_tensor)
+    return output_tensors
+
+
+def draw_e2e_res(dt_boxes, strs, img_path):
+    src_im = cv2.imread(img_path)
+    for box, str in zip(dt_boxes, strs):
+        box = box.astype(np.int32).reshape((-1, 1, 2))
+        cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
+        cv2.putText(
+            src_im,
+            str,
+            org=(int(box[0, 0, 0]), int(box[0, 0, 1])),
+            fontFace=cv2.FONT_HERSHEY_COMPLEX,
+            fontScale=0.7,
+            color=(0, 255, 0),
+            thickness=1,
+        )
+    return src_im
+
+
+def draw_text_det_res(dt_boxes, img):
+    for box in dt_boxes:
+        box = np.array(box).astype(np.int32).reshape(-1, 2)
+        cv2.polylines(img, [box], True, color=(255, 255, 0), thickness=2)
+    return img
+
+
+def resize_img(img, input_size=600):
+    """
+    resize img and limit the longest side of the image to input_size
+    """
+    img = np.array(img)
+    im_shape = img.shape
+    im_size_max = np.max(im_shape[0:2])
+    im_scale = float(input_size) / float(im_size_max)
+    img = cv2.resize(img, None, None, fx=im_scale, fy=im_scale)
+    return img
+
+
+def draw_ocr(
+    image,
+    boxes,
+    txts=None,
+    scores=None,
+    drop_score=0.5,
+    font_path="./doc/fonts/simfang.ttf",
+):
+    """
+    Visualize the results of OCR detection and recognition
+    args:
+        image(Image|array): RGB image
+        boxes(list): boxes with shape(N, 4, 2)
+        txts(list): the texts
+        scores(list): txxs corresponding scores
+        drop_score(float): only scores greater than drop_threshold will be visualized
+        font_path: the path of font which is used to draw text
+    return(array):
+        the visualized img
+    """
+    if scores is None:
+        scores = [1] * len(boxes)
+    box_num = len(boxes)
+    for i in range(box_num):
+        if scores is not None and (scores[i] < drop_score or math.isnan(scores[i])):
+            continue
+        box = np.reshape(np.array(boxes[i]), [-1, 1, 2]).astype(np.int64)
+        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
+    if txts is not None:
+        img = np.array(resize_img(image, input_size=600))
+        txt_img = text_visual(
+            txts,
+            scores,
+            img_h=img.shape[0],
+            img_w=600,
+            threshold=drop_score,
+            font_path=font_path,
+        )
+        img = np.concatenate([np.array(img), np.array(txt_img)], axis=1)
+        return img
+    return image
+
+
+def draw_ocr_box_txt(
+    image,
+    boxes,
+    txts=None,
+    scores=None,
+    drop_score=0.5,
+    font_path="./doc/fonts/simfang.ttf",
+):
+    h, w = image.height, image.width
+    img_left = image.copy()
+    img_right = np.ones((h, w, 3), dtype=np.uint8) * 255
+    random.seed(0)
+
+    draw_left = ImageDraw.Draw(img_left)
+    if txts is None or len(txts) != len(boxes):
+        txts = [None] * len(boxes)
+    for idx, (box, txt) in enumerate(zip(boxes, txts)):
+        if scores is not None and scores[idx] < drop_score:
+            continue
+        color = (random.randint(0, 255), random.randint(0, 255), random.randint(0, 255))
+        draw_left.polygon(box, fill=color)
+        img_right_text = draw_box_txt_fine((w, h), box, txt, font_path)
+        pts = np.array(box, np.int32).reshape((-1, 1, 2))
+        cv2.polylines(img_right_text, [pts], True, color, 1)
+        img_right = cv2.bitwise_and(img_right, img_right_text)
+    img_left = Image.blend(image, img_left, 0.5)
+    img_show = Image.new("RGB", (w * 2, h), (255, 255, 255))
+    img_show.paste(img_left, (0, 0, w, h))
+    img_show.paste(Image.fromarray(img_right), (w, 0, w * 2, h))
+    return np.array(img_show)
+
+
+def draw_box_txt_fine(img_size, box, txt, font_path="./doc/fonts/simfang.ttf"):
+    box_height = int(
+        math.sqrt((box[0][0] - box[3][0]) ** 2 + (box[0][1] - box[3][1]) ** 2)
+    )
+    box_width = int(
+        math.sqrt((box[0][0] - box[1][0]) ** 2 + (box[0][1] - box[1][1]) ** 2)
+    )
+
+    if box_height > 2 * box_width and box_height > 30:
+        img_text = Image.new("RGB", (box_height, box_width), (255, 255, 255))
+        draw_text = ImageDraw.Draw(img_text)
+        if txt:
+            font = create_font(txt, (box_height, box_width), font_path)
+            draw_text.text([0, 0], txt, fill=(0, 0, 0), font=font)
+        img_text = img_text.transpose(Image.ROTATE_270)
+    else:
+        img_text = Image.new("RGB", (box_width, box_height), (255, 255, 255))
+        draw_text = ImageDraw.Draw(img_text)
+        if txt:
+            font = create_font(txt, (box_width, box_height), font_path)
+            draw_text.text([0, 0], txt, fill=(0, 0, 0), font=font)
+
+    pts1 = np.float32(
+        [[0, 0], [box_width, 0], [box_width, box_height], [0, box_height]]
+    )
+    pts2 = np.array(box, dtype=np.float32)
+    M = cv2.getPerspectiveTransform(pts1, pts2)
+
+    img_text = np.array(img_text, dtype=np.uint8)
+    img_right_text = cv2.warpPerspective(
+        img_text,
+        M,
+        img_size,
+        flags=cv2.INTER_NEAREST,
+        borderMode=cv2.BORDER_CONSTANT,
+        borderValue=(255, 255, 255),
+    )
+    return img_right_text
+
+
+def create_font(txt, sz, font_path="./doc/fonts/simfang.ttf"):
+    font_size = int(sz[1] * 0.99)
+    font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
+    if int(PIL.__version__.split(".")[0]) < 10:
+        length = font.getsize(txt)[0]
+    else:
+        length = font.getlength(txt)
+
+    if length > sz[0]:
+        font_size = int(font_size * sz[0] / length)
+        font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
+    return font
+
+
+def str_count(s):
+    """
+    Count the number of Chinese characters,
+    a single English character and a single number
+    equal to half the length of Chinese characters.
+    args:
+        s(string): the input of string
+    return(int):
+        the number of Chinese characters
+    """
+    import string
+
+    count_zh = count_pu = 0
+    s_len = len(s)
+    en_dg_count = 0
+    for c in s:
+        if c in string.ascii_letters or c.isdigit() or c.isspace():
+            en_dg_count += 1
+        elif c.isalpha():
+            count_zh += 1
+        else:
+            count_pu += 1
+    return s_len - math.ceil(en_dg_count / 2)
+
+
+def text_visual(
+    texts, scores, img_h=400, img_w=600, threshold=0.0, font_path="./doc/simfang.ttf"
+):
+    """
+    create new blank img and draw txt on it
+    args:
+        texts(list): the text will be draw
+        scores(list|None): corresponding score of each txt
+        img_h(int): the height of blank img
+        img_w(int): the width of blank img
+        font_path: the path of font which is used to draw text
+    return(array):
+    """
+    if scores is not None:
+        assert len(texts) == len(
+            scores
+        ), "The number of txts and corresponding scores must match"
+
+    def create_blank_img():
+        blank_img = np.ones(shape=[img_h, img_w], dtype=np.int8) * 255
+        blank_img[:, img_w - 1 :] = 0
+        blank_img = Image.fromarray(blank_img).convert("RGB")
+        draw_txt = ImageDraw.Draw(blank_img)
+        return blank_img, draw_txt
+
+    blank_img, draw_txt = create_blank_img()
+
+    font_size = 20
+    txt_color = (0, 0, 0)
+    font = ImageFont.truetype(font_path, font_size, encoding="utf-8")
+
+    gap = font_size + 5
+    txt_img_list = []
+    count, index = 1, 0
+    for idx, txt in enumerate(texts):
+        index += 1
+        if scores[idx] < threshold or math.isnan(scores[idx]):
+            index -= 1
+            continue
+        first_line = True
+        while str_count(txt) >= img_w // font_size - 4:
+            tmp = txt
+            txt = tmp[: img_w // font_size - 4]
+            if first_line:
+                new_txt = str(index) + ": " + txt
+                first_line = False
+            else:
+                new_txt = "    " + txt
+            draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
+            txt = tmp[img_w // font_size - 4 :]
+            if count >= img_h // gap - 1:
+                txt_img_list.append(np.array(blank_img))
+                blank_img, draw_txt = create_blank_img()
+                count = 0
+            count += 1
+        if first_line:
+            new_txt = str(index) + ": " + txt + "   " + "%.3f" % (scores[idx])
+        else:
+            new_txt = "  " + txt + "  " + "%.3f" % (scores[idx])
+        draw_txt.text((0, gap * count), new_txt, txt_color, font=font)
+        # whether add new blank img or not
+        if count >= img_h // gap - 1 and idx + 1 < len(texts):
+            txt_img_list.append(np.array(blank_img))
+            blank_img, draw_txt = create_blank_img()
+            count = 0
+        count += 1
+    txt_img_list.append(np.array(blank_img))
+    if len(txt_img_list) == 1:
+        blank_img = np.array(txt_img_list[0])
+    else:
+        blank_img = np.concatenate(txt_img_list, axis=1)
+    return np.array(blank_img)
+
+
+def base64_to_cv2(b64str):
+    import base64
+
+    data = base64.b64decode(b64str.encode("utf8"))
+    data = np.frombuffer(data, np.uint8)
+    data = cv2.imdecode(data, cv2.IMREAD_COLOR)
+    return data
+
+
+def draw_boxes(image, boxes, scores=None, drop_score=0.5):
+    if scores is None:
+        scores = [1] * len(boxes)
+    for box, score in zip(boxes, scores):
+        if score < drop_score:
+            continue
+        box = np.reshape(np.array(box), [-1, 1, 2]).astype(np.int64)
+        image = cv2.polylines(np.array(image), [box], True, (255, 0, 0), 2)
+    return image
+
+
+def get_rotate_crop_image(img, points):
+    """
+    img_height, img_width = img.shape[0:2]
+    left = int(np.min(points[:, 0]))
+    right = int(np.max(points[:, 0]))
+    top = int(np.min(points[:, 1]))
+    bottom = int(np.max(points[:, 1]))
+    img_crop = img[top:bottom, left:right, :].copy()
+    points[:, 0] = points[:, 0] - left
+    points[:, 1] = points[:, 1] - top
+    """
+    assert len(points) == 4, "shape of points must be 4*2"
+    img_crop_width = int(
+        max(
+            np.linalg.norm(points[0] - points[1]), np.linalg.norm(points[2] - points[3])
+        )
+    )
+    img_crop_height = int(
+        max(
+            np.linalg.norm(points[0] - points[3]), np.linalg.norm(points[1] - points[2])
+        )
+    )
+    pts_std = np.float32(
+        [
+            [0, 0],
+            [img_crop_width, 0],
+            [img_crop_width, img_crop_height],
+            [0, img_crop_height],
+        ]
+    )
+    M = cv2.getPerspectiveTransform(points, pts_std)
+    dst_img = cv2.warpPerspective(
+        img,
+        M,
+        (img_crop_width, img_crop_height),
+        borderMode=cv2.BORDER_REPLICATE,
+        flags=cv2.INTER_CUBIC,
+    )
+    dst_img_height, dst_img_width = dst_img.shape[0:2]
+    if dst_img_height * 1.0 / dst_img_width >= 1.5:
+        dst_img = np.rot90(dst_img)
+    return dst_img
+
+
+def get_minarea_rect_crop(img, points):
+    bounding_box = cv2.minAreaRect(np.array(points).astype(np.int32))
+    points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0])
+
+    index_a, index_b, index_c, index_d = 0, 1, 2, 3
+    if points[1][1] > points[0][1]:
+        index_a = 0
+        index_d = 1
+    else:
+        index_a = 1
+        index_d = 0
+    if points[3][1] > points[2][1]:
+        index_b = 2
+        index_c = 3
+    else:
+        index_b = 3
+        index_c = 2
+
+    box = [points[index_a], points[index_b], points[index_c], points[index_d]]
+    crop_img = get_rotate_crop_image(img, np.array(box))
+    return crop_img
+
+
+# def check_gpu(use_gpu):
+#     if use_gpu and (
+#         not paddle.is_compiled_with_cuda() or paddle.device.get_device() == "cpu"
+#     ):
+#         use_gpu = False
+#     return use_gpu
+
+
+def _check_image_file(path):
+    img_end = {"jpg", "bmp", "png", "jpeg", "rgb", "tif", "tiff", "gif", "pdf"}
+    return any([path.lower().endswith(e) for e in img_end])
+
+
+def get_image_file_list(img_file, infer_list=None):
+    imgs_lists = []
+    if img_file is None or not os.path.exists(img_file):
+        raise Exception("not found any img file in {}".format(img_file))
+
+    if os.path.isfile(img_file) and _check_image_file(img_file):
+        imgs_lists.append(img_file)
+    elif os.path.isdir(img_file):
+        for single_file in os.listdir(img_file):
+            file_path = os.path.join(img_file, single_file)
+            if os.path.isfile(file_path) and _check_image_file(file_path):
+                imgs_lists.append(file_path)
+
+    if len(imgs_lists) == 0:
+        raise Exception("not found any img file in {}".format(img_file))
+    imgs_lists = sorted(imgs_lists)
+    return imgs_lists
+
+
+logger_initialized = {}
+@functools.lru_cache()
+def get_logger(name="ppocr", log_file=None, log_level=logging.DEBUG):
+    """Initialize and get a logger by name.
+    If the logger has not been initialized, this method will initialize the
+    logger by adding one or two handlers, otherwise the initialized logger will
+    be directly returned. During initialization, a StreamHandler will always be
+    added. If `log_file` is specified a FileHandler will also be added.
+    Args:
+        name (str): Logger name.
+        log_file (str | None): The log filename. If specified, a FileHandler
+            will be added to the logger.
+        log_level (int): The logger level. Note that only the process of
+            rank 0 is affected, and other processes will set the level to
+            "Error" thus be silent most of the time.
+    Returns:
+        logging.Logger: The expected logger.
+    """
+    logger = logging.getLogger(name)
+    if name in logger_initialized:
+        return logger
+    for logger_name in logger_initialized:
+        if name.startswith(logger_name):
+            return logger
+
+    formatter = logging.Formatter(
+        "[%(asctime)s] %(name)s %(levelname)s: %(message)s", datefmt="%Y/%m/%d %H:%M:%S"
+    )
+
+    stream_handler = logging.StreamHandler(stream=sys.stdout)
+    stream_handler.setFormatter(formatter)
+    logger.addHandler(stream_handler)
+    logger_initialized[name] = True
+    logger.propagate = False
+    return logger
+
+
+def get_rotate_crop_image(img, points):
+    """
+    img_height, img_width = img.shape[0:2]
+    left = int(np.min(points[:, 0]))
+    right = int(np.max(points[:, 0]))
+    top = int(np.min(points[:, 1]))
+    bottom = int(np.max(points[:, 1]))
+    img_crop = img[top:bottom, left:right, :].copy()
+    points[:, 0] = points[:, 0] - left
+    points[:, 1] = points[:, 1] - top
+    """
+    assert len(points) == 4, "shape of points must be 4*2"
+    img_crop_width = int(
+        max(
+            np.linalg.norm(points[0] - points[1]), np.linalg.norm(points[2] - points[3])
+        )
+    )
+    img_crop_height = int(
+        max(
+            np.linalg.norm(points[0] - points[3]), np.linalg.norm(points[1] - points[2])
+        )
+    )
+    pts_std = np.float32(
+        [
+            [0, 0],
+            [img_crop_width, 0],
+            [img_crop_width, img_crop_height],
+            [0, img_crop_height],
+        ]
+    )
+    M = cv2.getPerspectiveTransform(points, pts_std)
+    dst_img = cv2.warpPerspective(
+        img,
+        M,
+        (img_crop_width, img_crop_height),
+        borderMode=cv2.BORDER_REPLICATE,
+        flags=cv2.INTER_CUBIC,
+    )
+    dst_img_height, dst_img_width = dst_img.shape[0:2]
+    if dst_img_height * 1.0 / dst_img_width >= 1.5:
+        dst_img = np.rot90(dst_img)
+    return dst_img
+
+
+def get_minarea_rect_crop(img, points):
+    bounding_box = cv2.minAreaRect(np.array(points).astype(np.int32))
+    points = sorted(list(cv2.boxPoints(bounding_box)), key=lambda x: x[0])
+
+    index_a, index_b, index_c, index_d = 0, 1, 2, 3
+    if points[1][1] > points[0][1]:
+        index_a = 0
+        index_d = 1
+    else:
+        index_a = 1
+        index_d = 0
+    if points[3][1] > points[2][1]:
+        index_b = 2
+        index_c = 3
+    else:
+        index_b = 3
+        index_c = 2
+
+    box = [points[index_a], points[index_b], points[index_c], points[index_d]]
+    crop_img = get_rotate_crop_image(img, np.array(box))
+    return crop_img
+
+
+
+if __name__ == "__main__":
+    pass