fix comments and transform to transforms

configs/det/ch_ppocr_v1.1/ch_det_res18_db.yml

@@ -10,7 +10,7 @@ Global:
   # if pretrained_model is saved in static mode, load_static_weights must set to True
   load_static_weights: True
   cal_metric_during_train: False
-  pretrained_model: ./pretrain_models/MobileNetV3_large_x0_5_pretrained
+  pretrained_model: ./pretrain_models/ResNet18_vd_pretrained
   checkpoints: #./output/det_db_0.001_DiceLoss_256_pp_config_2.0b_4gpu/best_accuracy
   save_inference_dir:
   use_visualdl: False

ppocr/modeling/architectures/base_model.py

@@ -16,7 +16,7 @@ from __future__ import division
 from __future__ import print_function
 
 from paddle import nn
-from ppocr.modeling.transform import build_transform
+from ppocr.modeling.transforms import build_transform
 from ppocr.modeling.backbones import build_backbone
 from ppocr.modeling.necks import build_neck
 from ppocr.modeling.heads import build_head

ppocr/modeling/backbones/det_mobilenet_v3.py

@@ -111,6 +111,7 @@ class MobileNetV3(nn.Layer):
         i = 0
         inplanes = make_divisible(inplanes * scale)
         for (k, exp, c, se, nl, s) in cfg:
+            se = se and not self.disable_se
             if s == 2 and i > 2:
                 self.out_channels.append(inplanes)
                 self.stages.append(nn.Sequential(*block_list))
@@ -231,7 +232,7 @@ class ResidualUnit(nn.Layer):
             if_act=True,
             act=act,
             name=name + "_depthwise")
-        if self.if_se and not self.disable_se:
+        if self.if_se:
             self.mid_se = SEModule(mid_channels, name=name + "_se")
         self.linear_conv = ConvBNLayer(
             in_channels=mid_channels,
@@ -246,7 +247,7 @@ class ResidualUnit(nn.Layer):
     def forward(self, inputs):
         x = self.expand_conv(inputs)
         x = self.bottleneck_conv(x)
-        if self.if_se and not self.disable_se:
+        if self.if_se:
             x = self.mid_se(x)
         x = self.linear_conv(x)
         if self.if_shortcut:

ppocr/postprocess/db_postprocess.py

@@ -33,13 +33,14 @@ class DBPostProcess(object):
                  box_thresh=0.7,
                  max_candidates=1000,
                  unclip_ratio=2.0,
+                 use_dilation=False,
                  **kwargs):
         self.thresh = thresh
         self.box_thresh = box_thresh
         self.max_candidates = max_candidates
         self.unclip_ratio = unclip_ratio
         self.min_size = 3
-        self.dilation_kernel = np.array([[1, 1], [1, 1]])
+        self.dilation_kernel = None if not use_dilation else [[1, 1], [1, 1]]
 
     def boxes_from_bitmap(self, pred, _bitmap, dest_width, dest_height):
         '''
@@ -140,9 +141,12 @@ class DBPostProcess(object):
         boxes_batch = []
         for batch_index in range(pred.shape[0]):
             height, width = shape_list[batch_index]
-            mask = cv2.dilate(
-                np.array(segmentation[batch_index]).astype(np.uint8),
-                self.dilation_kernel)
+            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]
             boxes, scores = self.boxes_from_bitmap(pred[batch_index], mask,
                                                    width, height)