support new version of numpy (#8921)

* add finetune en doc & test=document_fix

* fix dead link & test=document_fix

* fix dead link & test=document_fix

* update check img

* fix det res dtype

* update args default type & test=document_fix

* fix numpy version

* support numpy1.24.0

applications/快速构建卡证类OCR.md

@@ -223,7 +223,7 @@ AIStudio项目链接：[快速构建卡证类OCR](https://aistudio.baidu.com/ais
 
 2）获取并解压预训练模型，如果要使用其他模型可以从模型库里自主选择合适模型。
 ```
-!wget -P work/pre_trained/   https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_distill_train.tar 
+!wget -P work/pre_trained/   https://paddleocr.bj.bcebos.com/PP-OCRv3/chinese/ch_PP-OCRv3_det_distill_train.tar
 !tar -vxf /home/aistudio/work/pre_trained/ch_PP-OCRv3_det_distill_train.tar -C /home/aistudio/work/pre_trained
 ```
 3） 安装必要依赖
@@ -275,7 +275,7 @@ AIStudio项目链接：[快速构建卡证类OCR](https://aistudio.baidu.com/ais
 
 ```python
 class DetLabelEncode(object):
-   
+
     # 修改检测标签的编码处，新增了参数分类数：num_classes，重写初始化方法，以及分类标签的读取
 
     def __init__(self, label_list, num_classes=8, **kwargs):
@@ -315,11 +315,11 @@ class DetLabelEncode(object):
                     classes.append(int(self.label_list.index(txt)))
 
         if len(boxes) == 0:
-            
+
             return None
         boxes = self.expand_points_num(boxes)
         boxes = np.array(boxes, dtype=np.float32)
-        txt_tags = np.array(txt_tags, dtype=np.bool)
+        txt_tags = np.array(txt_tags, dtype=np.bool_)
         classes = classes
         data['polys'] = boxes
         data['texts'] = txts
@@ -410,10 +410,10 @@ class MakeShrinkMap(object):
 
 
         data['shrink_map'] = gt
-       
+
         if self.num_classes > 1:
             data['class_mask'] = gt_class
-        
+
         data['shrink_mask'] = mask
         return data
 ```
@@ -634,10 +634,10 @@ class DBPostProcess(object):
         '''
         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)
+        xmin = np.clip(np.floor(box[:, 0].min()).astype(np.int32), 0, w - 1)
+        xmax = np.clip(np.ceil(box[:, 0].max()).astype(np.int32), 0, w - 1)
+        ymin = np.clip(np.floor(box[:, 1].min()).astype(np.int32), 0, h - 1)
+        ymax = np.clip(np.ceil(box[:, 1].max()).astype(np.int32), 0, h - 1)
 
         mask = np.zeros((ymax - ymin + 1, xmax - xmin + 1), dtype=np.uint8)
         box[:, 0] = box[:, 0] - xmin
@@ -752,11 +752,11 @@ class DBPostProcess(object):
 其他命令：
 ```
 !python /home/aistudio/work/PaddleOCR/tools/eval.py  -c  /home/aistudio/work/PaddleOCR/configs/det/det_mv3_db.yml
-!python /home/aistudio/work/PaddleOCR/tools/infer_det.py  -c  /home/aistudio/work/PaddleOCR/configs/det/det_mv3_db.yml 
+!python /home/aistudio/work/PaddleOCR/tools/infer_det.py  -c  /home/aistudio/work/PaddleOCR/configs/det/det_mv3_db.yml
 ```
 模型推理
 ```
-!python /home/aistudio/work/PaddleOCR/tools/infer/predict_det.py --image_dir="/home/aistudio/work/test_img/" --det_model_dir="/home/aistudio/work/PaddleOCR/output/infer" 
+!python /home/aistudio/work/PaddleOCR/tools/infer/predict_det.py --image_dir="/home/aistudio/work/test_img/" --det_model_dir="/home/aistudio/work/PaddleOCR/output/infer"
 ```
 
 ## 5 总结

deploy/avh/convert_image.py

@@ -24,6 +24,7 @@ import math
 from PIL import Image
 import numpy as np
 
+
 def resize_norm_img(img, image_shape, padding=True):
     imgC, imgH, imgW = image_shape
     h = img.shape[0]
@@ -61,9 +62,8 @@ def create_header_file(name, tensor_name, tensor_data, output_path):
     raw_path = file_path.with_suffix(".h").resolve()
     with open(raw_path, "w") as header_file:
         header_file.write(
-            "\n"
-            + f"const size_t {tensor_name}_len = {tensor_data.size};\n"
-            + f'__attribute__((section(".data.tvm"), aligned(16))) float {tensor_name}[] = '
+            "\n" + f"const size_t {tensor_name}_len = {tensor_data.size};\n" +
+            f'__attribute__((section(".data.tvm"), aligned(16))) float {tensor_name}[] = '
         )
 
         header_file.write("{")
@@ -80,22 +80,21 @@ def create_headers(image_name):
 
     # Resize image to 32x320
     img = cv2.imread(img_path)
-    img = resize_norm_img(img, [3,32,320])
+    img = resize_norm_img(img, [3, 32, 320])
     img_data = img.astype("float32")
-    
+
     # # Add the batch dimension, as we are expecting 4-dimensional input: NCHW.
     img_data = np.expand_dims(img_data, axis=0)
 
     # Create input header file
     create_header_file("inputs", "input", img_data, "./include")
     # Create output header file
-    output_data = np.zeros([7760], np.float)
+    output_data = np.zeros([7760], np.float32)
     create_header_file(
         "outputs",
         "output",
         output_data,
-        "./include",
-    )
+        "./include", )
 
 
 if __name__ == "__main__":

deploy/hubserving/ocr_det/module.py

@@ -122,7 +122,7 @@ class OCRDet(hub.Module):
             rec_res_final = []
             for dno in range(len(dt_boxes)):
                 rec_res_final.append({
-                    'text_region': dt_boxes[dno].astype(np.int).tolist()
+                    'text_region': dt_boxes[dno].astype(np.int32).tolist()
                 })
             all_results.append(rec_res_final)
         return all_results

deploy/hubserving/ocr_system/module.py

@@ -130,7 +130,7 @@ class OCRSystem(hub.Module):
                 rec_res_final.append({
                     'text': text,
                     'confidence': float(score),
-                    'text_region': dt_boxes[dno].astype(np.int).tolist()
+                    'text_region': dt_boxes[dno].astype(np.int32).tolist()
                 })
             all_results.append(rec_res_final)
         return all_results

ppocr/data/imaug/abinet_aug.py

@@ -205,7 +205,7 @@ class CVRandomAffine(object):
                      for x, y in startpoints]
 
         rect = cv2.minAreaRect(np.array(endpoints))
-        bbox = cv2.boxPoints(rect).astype(dtype=np.int)
+        bbox = cv2.boxPoints(rect).astype(dtype=np.int32)
         max_x, max_y = bbox[:, 0].max(), bbox[:, 1].max()
         min_x, min_y = bbox[:, 0].min(), bbox[:, 1].min()
 
@@ -234,9 +234,9 @@ class CVRandomPerspective(object):
 
     def get_params(self, width, height, distortion):
         offset_h = sample_asym(
-            distortion * height / 2, size=4).astype(dtype=np.int)
+            distortion * height / 2, size=4).astype(dtype=np.int32)
         offset_w = sample_asym(
-            distortion * width / 2, size=4).astype(dtype=np.int)
+            distortion * width / 2, size=4).astype(dtype=np.int32)
         topleft = (offset_w[0], offset_h[0])
         topright = (width - 1 - offset_w[1], offset_h[1])
         botright = (width - 1 - offset_w[2], height - 1 - offset_h[2])
@@ -256,7 +256,7 @@ class CVRandomPerspective(object):
 
         # TODO: more robust way to crop image
         rect = cv2.minAreaRect(endpoints)
-        bbox = cv2.boxPoints(rect).astype(dtype=np.int)
+        bbox = cv2.boxPoints(rect).astype(dtype=np.int32)
         max_x, max_y = bbox[:, 0].max(), bbox[:, 1].max()
         min_x, min_y = bbox[:, 0].min(), bbox[:, 1].min()
         min_x, min_y = max(min_x, 0), max(min_y, 0)
@@ -441,7 +441,8 @@ class SVTRGeometry(object):
         self.p = p
         self.transforms = []
         self.transforms.append(CVRandomRotation(degrees=degrees))
-        self.transforms.append(CVRandomAffine(
+        self.transforms.append(
+            CVRandomAffine(
                 degrees=degrees, translate=translate, scale=scale, shear=shear))
         self.transforms.append(CVRandomPerspective(distortion=distortion))
 
@@ -455,4 +456,4 @@ class SVTRGeometry(object):
                 img = self.transforms[random.randint(0, 2)](img)
             return img
         else:
-            return img
\ No newline at end of file
+            return img

ppocr/data/imaug/fce_aug.py

@@ -208,7 +208,7 @@ class RandomCropFlip:
         for polygon in all_polys:
             rect = cv2.minAreaRect(polygon.astype(np.int32).reshape(-1, 2))
             box = cv2.boxPoints(rect)
-            box = np.int0(box)
+            box = np.int64(box)
             text_polys.append([box[0], box[1], box[2], box[3]])
 
         polys = np.array(text_polys, dtype=np.int32)

ppocr/data/imaug/fce_targets.py

@@ -22,10 +22,12 @@ from numpy.fft import fft
 from numpy.linalg import norm
 import sys
 
+
 def vector_slope(vec):
     assert len(vec) == 2
     return abs(vec[1] / (vec[0] + 1e-8))
 
+
 class FCENetTargets:
     """Generate the ground truth targets of FCENet: Fourier Contour Embedding
     for Arbitrary-Shaped Text Detection.
@@ -107,7 +109,9 @@ class FCENetTargets:
         for i in range(1, n):
             current_line_len = i * delta_length
 
-            while current_edge_ind + 1 < len(length_cumsum) and current_line_len >= length_cumsum[current_edge_ind + 1]:
+            while current_edge_ind + 1 < len(
+                    length_cumsum) and current_line_len >= length_cumsum[
+                        current_edge_ind + 1]:
                 current_edge_ind += 1
 
             current_edge_end_shift = current_line_len - length_cumsum[
@@ -239,10 +243,9 @@ class FCENetTargets:
             head_inds = [head_start, head_end]
             tail_inds = [tail_start, tail_end]
         else:
-            if vector_slope(points[1] - points[0]) + vector_slope(
-                    points[3] - points[2]) < vector_slope(points[
-                        2] - points[1]) + vector_slope(points[0] - points[
-                            3]):
+            if vector_slope(points[1] - points[0]) + vector_slope(points[
+                    3] - points[2]) < vector_slope(points[2] - points[
+                        1]) + vector_slope(points[0] - points[3]):
                 horizontal_edge_inds = [[0, 1], [2, 3]]
                 vertical_edge_inds = [[3, 0], [1, 2]]
             else:
@@ -582,7 +585,7 @@ class FCENetTargets:
         lv_ignore_polys = [[] for i in range(len(lv_size_divs))]
         level_maps = []
         for poly in text_polys:
-            polygon = np.array(poly, dtype=np.int).reshape((1, -1, 2))
+            polygon = np.array(poly, dtype=np.int32).reshape((1, -1, 2))
             _, _, box_w, box_h = cv2.boundingRect(polygon)
             proportion = max(box_h, box_w) / (h + 1e-8)
 
@@ -591,7 +594,7 @@ class FCENetTargets:
                     lv_text_polys[ind].append(poly / lv_size_divs[ind])
 
         for ignore_poly in ignore_polys:
-            polygon = np.array(ignore_poly, dtype=np.int).reshape((1, -1, 2))
+            polygon = np.array(ignore_poly, dtype=np.int32).reshape((1, -1, 2))
             _, _, box_w, box_h = cv2.boundingRect(polygon)
             proportion = max(box_h, box_w) / (h + 1e-8)
 

ppocr/data/imaug/label_ops.py

@@ -64,7 +64,7 @@ class DetLabelEncode(object):
             return None
         boxes = self.expand_points_num(boxes)
         boxes = np.array(boxes, dtype=np.float32)
-        txt_tags = np.array(txt_tags, dtype=np.bool)
+        txt_tags = np.array(txt_tags, dtype=np.bool_)
 
         data['polys'] = boxes
         data['texts'] = txts
@@ -218,7 +218,7 @@ class E2ELabelEncodeTest(BaseRecLabelEncode):
             else:
                 txt_tags.append(False)
         boxes = np.array(boxes, dtype=np.float32)
-        txt_tags = np.array(txt_tags, dtype=np.bool)
+        txt_tags = np.array(txt_tags, dtype=np.bool_)
         data['polys'] = boxes
         data['ignore_tags'] = txt_tags
         temp_texts = []
@@ -254,7 +254,7 @@ class E2ELabelEncodeTrain(object):
             else:
                 txt_tags.append(False)
         boxes = np.array(boxes, dtype=np.float32)
-        txt_tags = np.array(txt_tags, dtype=np.bool)
+        txt_tags = np.array(txt_tags, dtype=np.bool_)
 
         data['polys'] = boxes
         data['texts'] = txts

ppocr/modeling/heads/proposal_local_graph.py

@@ -40,7 +40,7 @@ def fill_hole(input_mask):
     mask = np.zeros((h + 4, w + 4), np.uint8)
 
     cv2.floodFill(canvas, mask, (0, 0), 1)
-    canvas = canvas[1:h + 1, 1:w + 1].astype(np.bool)
+    canvas = canvas[1:h + 1, 1:w + 1].astype(np.bool_)
 
     return ~canvas | input_mask
 

ppocr/postprocess/drrg_postprocess.py

@@ -68,7 +68,7 @@ def graph_propagation(edges, scores, text_comps, edge_len_thr=50.):
             score_dict[edge[0], edge[1]] = scores[i]
 
     nodes = np.sort(np.unique(edges.flatten()))
-    mapping = -1 * np.ones((np.max(nodes) + 1), dtype=np.int)
+    mapping = -1 * np.ones((np.max(nodes) + 1), dtype=np.int32)
     mapping[nodes] = np.arange(nodes.shape[0])
     order_inds = mapping[edges]
     vertices = [Node(node) for node in nodes]
@@ -93,9 +93,8 @@ def connected_components(nodes, score_dict, link_thr):
         while node_queue:
             node = node_queue.pop(0)
             neighbors = set([
-                neighbor for neighbor in node.links
-                if score_dict[tuple(sorted([node.ind, neighbor.ind]))] >=
-                link_thr
+                neighbor for neighbor in node.links if
+                score_dict[tuple(sorted([node.ind, neighbor.ind]))] >= link_thr
             ])
             neighbors.difference_update(cluster)
             nodes.difference_update(neighbors)

ppocr/postprocess/fce_postprocess.py

@@ -31,7 +31,7 @@ def fill_hole(input_mask):
     mask = np.zeros((h + 4, w + 4), np.uint8)
 
     cv2.floodFill(canvas, mask, (0, 0), 1)
-    canvas = canvas[1:h + 1, 1:w + 1].astype(np.bool)
+    canvas = canvas[1:h + 1, 1:w + 1].astype(np.bool_)
 
     return ~canvas | input_mask
 
@@ -234,7 +234,7 @@ class FCEPostProcess(object):
                 poly = np.array(boundary[:-1]).reshape(-1, 2).astype(np.float32)
                 score = boundary[-1]
                 points = cv2.boxPoints(cv2.minAreaRect(poly))
-                points = np.int0(points)
+                points = np.int64(points)
                 new_boundaries.append(points.reshape(-1).tolist() + [score])
                 boundaries = new_boundaries