!12517 enhance for ctpn performance

From: @qujianwei
Reviewed-by: @c_34,@oacjiewen
Signed-off-by: @c_34

model_zoo/official/cv/ctpn/src/CTPN/anchor_generator.py

@@ -12,10 +12,10 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ============================================================================
-"""FasterRcnn anchor generator."""
+"""CTPN anchor generator."""
 import numpy as np
 class AnchorGenerator():
-    """Anchor generator for FasterRcnn."""
+    """Anchor generator for CTPN."""
     def __init__(self, config):
         """Anchor generator init method."""
         self.base_size = config.anchor_base

model_zoo/official/cv/ctpn/src/CTPN/bbox_assign_sample.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ============================================================================
-"""FasterRcnn positive and negative sample screening for RPN."""
+"""CTPN positive and negative sample screening for RPN."""
 
 import numpy as np
 import mindspore.nn as nn

model_zoo/official/cv/ctpn/src/CTPN/proposal_generator.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ============================================================================
-"""FasterRcnn proposal generator."""
+"""CTPN proposal generator."""
 
 import numpy as np
 import mindspore.nn as nn

model_zoo/official/cv/ctpn/src/CTPN/rpn.py

@@ -49,25 +49,24 @@ class RpnRegClsBlock(nn.Cell):
         self.lstm_fc = nn.Dense(2*config.hidden_size, 512).to_float(mstype.float16)
         self.rpn_cls = nn.Dense(in_channels=512, out_channels=num_anchors * cls_out_channels).to_float(mstype.float16)
         self.rpn_reg = nn.Dense(in_channels=512, out_channels=num_anchors * 4).to_float(mstype.float16)
-        self.shape1 = (config.num_step, config.rnn_batch_size, -1)
+        self.shape1 = (-1, config.num_step, config.rnn_batch_size)
-        self.shape2 = (-1, config.batch_size, config.rnn_batch_size, config.num_step)
+        self.shape2 = (config.batch_size, -1, config.rnn_batch_size, config.num_step)
         self.transpose = P.Transpose()
         self.print = P.Print()
-        self.dropout = nn.Dropout(0.8)
 
     def construct(self, x):
         x = self.reshape(x, self.shape)
         x = self.lstm_fc(x)
         x1 = self.rpn_cls(x)
+        x1 = self.transpose(x1, (1, 0))
         x1 = self.reshape(x1, self.shape1)
-        x1 = self.transpose(x1, (2, 1, 0))
+        x1 = self.transpose(x1, (0, 2, 1))
         x1 = self.reshape(x1, self.shape2)
-        x1 = self.transpose(x1, (1, 0, 2, 3))
         x2 = self.rpn_reg(x)
+        x2 = self.transpose(x2, (1, 0))
         x2 = self.reshape(x2, self.shape1)
-        x2 = self.transpose(x2, (2, 1, 0))
+        x2 = self.transpose(x2, (0, 2, 1))
         x2 = self.reshape(x2, self.shape2)
-        x2 = self.transpose(x2, (1, 0, 2, 3))
         return x1, x2
 
 class RPN(nn.Cell):

model_zoo/official/cv/ctpn/src/dataset.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 # ============================================================================
 
-"""FasterRcnn dataset"""
+"""CTPN dataset"""
 from __future__ import division
 import os
 import numpy as np

model_zoo/official/cv/ctpn/src/network_define.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ============================================================================
-"""FasterRcnn training network wrapper."""
+"""CTPN training network wrapper."""
 
 import time
 import numpy as np
@@ -82,7 +82,7 @@ class LossCallBack(Callback):
             loss_file.close()
 
 class LossNet(nn.Cell):
-    """FasterRcnn loss method"""
+    """CTPN loss method"""
     def construct(self, x1, x2, x3):
         return x1