Skip to content

📄 Awesome OCR multiple programing languages toolkits based on ONNX Runtime, OpenVINO, MNN, PaddlePaddle, TensorRT and PyTorch.

add ch_ppocr_mobile_v1.1_det module

SWHLa417efbb2843
diff --git a/python/ch_ppocr_mobile_v1.1_det/demo.py b/python/ch_ppocr_mobile_v1.1_det/demo.py
new file mode 100644
index 0000000..94a75a0
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/demo.py
@@ -0,0 +1,17 @@
+# !/usr/bin/env python
+# -*- encoding: utf-8 -*-
+# @File: demo.py
+# @Time: 2021/01/16 16:33:49
+# @Author: Max
+from text_detect import TextDetector
+import cv2
+
+test_detector = TextDetector()
+
+image_path = 'images/1.jpg'
+
+# dst_boxes: 检测到图像中的文本框坐标,ndarray格式
+# (10, 4, 2)→[10个,4个坐标,每个坐标两个点]
+dst_boxes, im = test_detector(image_path)
+
+cv2.imwrite('images/det_results.jpg', im)
\ No newline at end of file
diff --git a/python/ch_ppocr_mobile_v1.1_det/images/1.jpg b/python/ch_ppocr_mobile_v1.1_det/images/1.jpg
new file mode 100644
index 0000000..d352416
Binary files /dev/null and b/python/ch_ppocr_mobile_v1.1_det/images/1.jpg differ
diff --git a/python/ch_ppocr_mobile_v1.1_det/images/det_results.jpg b/python/ch_ppocr_mobile_v1.1_det/images/det_results.jpg
new file mode 100644
index 0000000..b37e216
Binary files /dev/null and b/python/ch_ppocr_mobile_v1.1_det/images/det_results.jpg differ
diff --git a/python/ch_ppocr_mobile_v1.1_det/requirements.txt b/python/ch_ppocr_mobile_v1.1_det/requirements.txt
new file mode 100644
index 0000000..19d444b
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/requirements.txt
@@ -0,0 +1,6 @@
+onnxruntime
+opencv-python
+numpy
+pyclipper
+Shapely
+python==3.6.4
\ No newline at end of file
diff --git a/python/ch_ppocr_mobile_v1.1_det/text_detect/__init__.py b/python/ch_ppocr_mobile_v1.1_det/text_detect/__init__.py
new file mode 100644
index 0000000..2697968
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/text_detect/__init__.py
@@ -0,0 +1 @@
+from .text_detect import TextDetector
\ No newline at end of file
diff --git a/python/ch_ppocr_mobile_v1.1_det/text_detect/text_detect.py b/python/ch_ppocr_mobile_v1.1_det/text_detect/text_detect.py
new file mode 100644
index 0000000..495fa3a
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/text_detect/text_detect.py
@@ -0,0 +1,59 @@
+# !/usr/bin/env python
+# -*- encoding: utf-8 -*-
+# @File: get_text_location.py
+# @Time: 2021/01/16 11:56:04
+# @Author: Max
+from pathlib import Path
+import onnxruntime
+import time
+import argparse
+
+import cv2
+from .utils import DBPostProcess, DBPreProcess, utils
+
+root_path = Path(__file__).parent
+det_model_path = str(root_path / 'weights' / 'ch_mobile_v1.1_det.onnx')
+
+
+class TextDetector(object):
+ def __init__(self):
+ self.preprocess_op = DBPreProcess()
+ self.session = onnxruntime.InferenceSession(det_model_path)
+ self.postprocess_op = DBPostProcess()
+
+ @staticmethod
+ def load_image(image_path):
+ img, flag = utils.check_and_read_gif(image_path)
+ if not flag:
+ img = cv2.imread(image_path)
+ if img is None:
+ raise ValueError('图像有问题')
+ return img
+
+ def __call__(self, image_path):
+ img = self.load_image(image_path)
+ ori_im = img.copy()
+ im, ratio_list = self.preprocess_op(img)
+ if im is None:
+ return None, 0
+
+ inputs = {self.session.get_inputs()[0].name: im}
+ outputs = self.session.run(None, inputs)[0]
+
+ dt_boxes = self.postprocess_op(outputs, [ratio_list], ori_im.shape)
+ plot_img = utils.draw_text_det_res(dt_boxes, img)
+ return dt_boxes, plot_img
+
+
+if __name__ == "__main__":
+ parser = argparse.ArgumentParser()
+ parser.add_argument('--image_path', type=str,
+ default='images/1.jpg')
+ args = parser.parse_args()
+
+ text_detector = TextDetector()
+ image_path = r'images/2021-01-16_11-39-35.png'
+ dt_boxes, plot_img = text_detector(args.image_path)
+ cv2.imwrite(f'det_results.jpg', plot_img)
+
+
diff --git a/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/__init__.py b/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/__init__.py
new file mode 100644
index 0000000..132519d
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/__init__.py
@@ -0,0 +1,7 @@
+# !/usr/bin/env python
+# -*- encoding: utf-8 -*-
+# @File: __init__.py
+# @Time: 2021/01/16 12:00:44
+# @Author: Max
+from .db_process import *
+from .utils import *
\ No newline at end of file
diff --git a/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/db_process.py b/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/db_process.py
new file mode 100644
index 0000000..1259590
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/db_process.py
@@ -0,0 +1,277 @@
+# !/usr/bin/env python
+# -*- encoding: utf-8 -*-
+# @File: db_process.py
+# @Time: 2021/01/16 12:00:18
+# @Author: Max
+import sys
+from copy import deepcopy
+
+import cv2
+import numpy as np
+import pyclipper
+from shapely.geometry import Polygon
+
+
+class DBPreProcess(object):
+ def __init__(self):
+ super(DBPreProcess, self).__init__()
+ self.max_side_len = 2400
+
+ def resize_image(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)
+ if resize_h % 32 == 0:
+ resize_h = resize_h
+ elif resize_h // 32 <= 1:
+ resize_h = 32
+ else:
+ resize_h = (resize_h // 32) * 32
+ if resize_w % 32 == 0:
+ resize_w = resize_w
+ elif resize_w // 32 <= 1:
+ resize_w = 32
+ else:
+ resize_w = (resize_w // 32) * 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 normalize(self, im):
+ """
+ Normalize image
+ :param im: input image
+ :return: Normalized image
+ """
+ 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[:, :, 0] -= img_mean[0]
+ im[:, :, 1] -= img_mean[1]
+ im[:, :, 2] -= img_mean[2]
+ im[:, :, 0] /= img_std[0]
+ im[:, :, 1] /= img_std[1]
+ im[:, :, 2] /= img_std[2]
+ channel_swap = (2, 0, 1)
+ im = im.transpose(channel_swap)
+ return im
+
+ def __call__(self, im):
+ im, (ratio_h, ratio_w) = self.resize_image(im)
+ im = self.normalize(im)
+ im = im[np.newaxis, :]
+ return [im, (ratio_h, ratio_w)]
+
+
+class DBPostProcess(object):
+ def __init__(self):
+ self.thresh = 0.3
+ self.box_thresh = 0.5
+ self.max_candidates = 1000
+ self.unclip_ratio = 1.6
+ self.min_size = 3
+ self.dilation_kernel = np.array([[1, 1], [1, 1]])
+
+ def boxes_from_bitmap(self, pred, mask):
+ """
+ Get boxes from the binarized image predicted by DB.
+ :param pred: the binarized image predicted by DB.
+ :param mask: new 'pred' after threshold filtering.
+ :return: (boxes, the score of each boxes)
+ """
+ dest_height, dest_width = pred.shape[-2:]
+ bitmap = deepcopy(mask)
+ 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 = np.zeros((num_contours, 4, 2), dtype=np.int16)
+ scores = np.zeros((num_contours, ), dtype=np.float32)
+
+ 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)
+ score = self.box_score_fast(pred, points.reshape(-1, 2))
+ if self.box_thresh > score:
+ continue
+
+ box = self.unclip(points).reshape(-1, 1, 2)
+ box, sside = self.get_mini_boxes(box)
+ if sside < self.min_size + 2:
+ continue
+ box = np.array(box)
+ if not isinstance(dest_width, int):
+ dest_width = dest_width.item()
+ dest_height = dest_height.item()
+
+ 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[index, :, :] = box.astype(np.int16)
+ scores[index] = score
+ return boxes, scores
+
+ def unclip(self, box):
+ """
+ Shrink or expand the boxaccording to 'unclip_ratio'
+ :param box: The predicted box.
+ :return: uncliped box
+ """
+ unclip_ratio = self.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):
+ """
+ Get boxes from the contour or box.
+ :param contour: The predicted contour.
+ :return: The predicted box.
+ """
+ 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):
+ """
+ Calculate the score of box.
+ :param bitmap: The binarized image predicted by DB.
+ :param _box: The predicted box
+ :return: score
+ """
+ h, w = bitmap.shape[:2]
+ box = _box.copy()
+ xmin = np.clip(np.floor(box[:, 0].min()).astype(np.int), 0, w - 1)
+ xmax = np.clip(np.ceil(box[:, 0].max()).astype(np.int), 0, w - 1)
+ ymin = np.clip(np.floor(box[:, 1].min()).astype(np.int), 0, h - 1)
+ ymax = np.clip(np.ceil(box[:, 1].max()).astype(np.int), 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(np.int32), 1)
+ return cv2.mean(bitmap[ymin:ymax + 1, xmin:xmax + 1], mask)[0]
+
+ @staticmethod
+ def order_points_clockwise(pts):
+ """
+ reference from: https://github.com/jrosebr1/imutils/blob/master/imutils/perspective.py
+ # sort the points based on their x-coordinates
+ """
+ x_sorted = pts[np.argsort(pts[:, 0]), :]
+
+ # grab the left-most and right-most points from the sorted
+ # x-roodinate points
+ left_most = x_sorted[:2, :]
+ right_most = x_sorted[2:, :]
+
+ # now, sort the left-most coordinates according to their
+ # y-coordinates so we can grab the top-left and bottom-left
+ # points, respectively
+ left_most = left_most[np.argsort(left_most[:, 1]), :]
+ (tl, bl) = left_most
+
+ right_most = right_most[np.argsort(right_most[:, 1]), :]
+ (tr, br) = right_most
+
+ rect = np.array([tl, tr, br, bl], dtype="float32")
+ return rect
+
+ @staticmethod
+ def clip_det_res(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:
+ box = self.order_points_clockwise(box)
+ box = self.clip_det_res(box, img_height, img_width)
+ dt_boxes_new.append(box)
+ return dt_boxes
+
+ def __call__(self, pred, ratio_list, ori_im_shape):
+ pred = pred[:, 0, :, :]
+ segmentation = pred > self.thresh
+ boxes_batch = []
+ for batch_index in range(pred.shape[0]):
+
+ mask = cv2.dilate(
+ np.array(segmentation[batch_index]).astype(np.uint8),
+ self.dilation_kernel)
+ tmp_boxes, tmp_scores = self.boxes_from_bitmap(pred[batch_index],
+ mask)
+ boxes = []
+ for k in range(len(tmp_boxes)):
+ if tmp_scores[k] > self.box_thresh:
+ boxes.append(tmp_boxes[k])
+ if len(boxes) > 0:
+ boxes = np.array(boxes)
+ ratio_h, ratio_w = ratio_list[batch_index]
+ boxes[:, :, 0] = boxes[:, :, 0] / ratio_w
+ boxes[:, :, 1] = boxes[:, :, 1] / ratio_h
+ boxes_batch.append(boxes)
+ dt_boxes = self.filter_tag_det_res(boxes_batch[0], ori_im_shape)
+ return dt_boxes
diff --git a/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/utils.py b/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/utils.py
new file mode 100644
index 0000000..53f5533
--- /dev/null
+++ b/python/ch_ppocr_mobile_v1.1_det/text_detect/utils/utils.py
@@ -0,0 +1,37 @@
+# !/usr/bin/env python
+# -*- encoding: utf-8 -*-
+# @File: utils.py
+# @Time: 2021/01/16 12:00:01
+# @Author: Max
+import copy
+from pathlib import Path
+
+import cv2
+import numpy as np
+
+
+def check_and_read_gif(img_path):
+ if Path(img_path).suffix in ['.gif', '.GIF']:
+ gif = cv2.VideoCapture(img_path)
+ ret, frame = gif.read()
+ if not ret:
+ print(f"Cannot read {str(img_path)}. This gif image maybe corrupted.")
+ return None, False
+ if len(frame.shape) == 2 or frame.shape[-1] == 1:
+ frame = cv2.cvtColor(frame, cv2.COLOR_GRAY2RGB)
+ imgvalue = frame[:, :, ::-1]
+ return imgvalue, True
+ return None, False
+
+
+def draw_text_det_res(dt_boxes, raw_im):
+ src_im = copy.deepcopy(raw_im)
+ for i, box in enumerate(dt_boxes):
+ box = np.array(box).astype(np.int32).reshape(-1, 2)
+ cv2.polylines(src_im, [box], True,
+ color=(255, 255, 0),
+ thickness=2)
+ cv2.putText(src_im, str(i), (int(box[0][0]), int(box[0][1])),
+ cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 3)
+ return src_im
+
diff --git a/python/ch_ppocr_mobile_v1.1_det/text_detect/weights/ch_mobile_v1.1_det.onnx b/python/ch_ppocr_mobile_v1.1_det/text_detect/weights/ch_mobile_v1.1_det.onnx
new file mode 100644
index 0000000..35131f7
Binary files /dev/null and b/python/ch_ppocr_mobile_v1.1_det/text_detect/weights/ch_mobile_v1.1_det.onnx differ
diff --git a/python/ch_ppocr_mobile_v1.1_det/whl/text_detect-1.0.0-py3-none-any.whl b/python/ch_ppocr_mobile_v1.1_det/whl/text_detect-1.0.0-py3-none-any.whl
new file mode 100644
index 0000000..e8fb79c
Binary files /dev/null and b/python/ch_ppocr_mobile_v1.1_det/whl/text_detect-1.0.0-py3-none-any.whl differ
diff --git a/python/readme.md b/python/readme.md
new file mode 100644
index 0000000..2134ef8
--- /dev/null
+++ b/python/readme.md
@@ -0,0 +1,32 @@
+## 通用超轻量文本检测模型 ch_ppocr_mobile_v1.1_det
+**Note: 支持任意尺寸输入**
+#### 第一种使用方法:
+- 直接安装提供的whl包
+```shell
+pip install text_detect-1.0.0-py3-none-any.whl
+```
+- 运行
+```python
+from text_detect import TextDetector
+import cv2
+
+test_detector = TextDetector()
+
+image_path = 'images/1.jpg'
+
+# dst_boxes: 检测到图像中的文本框坐标,ndarray格式,
+# e.g (10, 4, 2)→[10个,4个坐标对,每个坐标两个点]
+dst_boxes, im = test_detector(image_path)
+
+cv2.imwrite('det_results.jpg', im)
+```
+#### 第二种使用方法
+- 下载`ch_ppocr_mobile_v1.1_det`整个文件夹到本地目录
+- 运行
+```shell
+cd ch_mobile_det
+python demo.py
+```
+- 即可看到检测结果图(`images/det_results.jpg`),效果如下:
+
+![det_result](./ch_ppocr_mobile_v1.1_det/images/det_results.jpg)
\ No newline at end of file
diff --git a/python/readme.txt b/python/readme.txt
deleted file mode 100644
index b3a4252..0000000
--- a/python/readme.txt
+++ /dev/null
@@ -1 +0,0 @@
-placeholder
\ No newline at end of file