修改python代码中的预处理过程

Doc/Tutorial_Python/Unet.md

@@ -14,23 +14,27 @@
 
 ​      1.尺度变换，将图像resize到256x256大小
 
-​      2.归一化，将数据归一化到[0.0, 1.0]之间
+​      2.数据排布，将数据从HWC转换为CHW，再将维度转换为NCHW
 
-​      3.数据排布，将数据从HWC转换为CHW，再将维度转换为NCHW
+​      3.归一化，将数据归一化到[0.0, 1.0]之间
 
 本示例代码通过如下方式实现预处理操作：
 
 ```python
 def Preprocessing(pil_img, newW, newH):
     assert newW > 0 and newH > 0, 'Scale is too small'  
-    img_nd = cv2.resize(pil_img, (newW, newH))          # 将图像尺寸修改为256x256
-    img_nd = cv2.cvtColor(img_nd, cv2.COLOR_BGR2RGB)    # BGR转换为RGB
-
-    img_trans = img_nd.transpose((2, 0, 1))             # HWC转换为CHW
-    if img_trans.max() > 1:                             # 保证数据处于0-1之间的浮点数
-        img_trans = img_trans / 255.0  
-    img_trans = np.expand_dims(img_trans, 0)            # CHW转换为NCHW
-    img = img_trans.astype(np.float32)                  # 转换成浮点型数据
+    img_nd = cv2.cvtColor(pil_img, cv2.COLOR_BGR2RGB)    # BGR转换为RGB
+    img_nd = cv2.resize(img_nd, (newW, newH))            # 将图像尺寸修改为256x256
+    
+    if len(img_nd.shape) == 2:
+        img_nd = np.expand_dims(img_nd, axis=2)
+
+    img_trans = img_nd.transpose((2, 0, 1))              # HWC转换为CHW
+    img_trans = np.expand_dims(img_trans, 0)             # CHW扩展为NCHW
+    img_trans = np.ascontiguousarray(img_trans)          # 保证内存连续存储
+    img_trans = img_trans.astype(np.float32)             # 转换成浮点型数据
+    if img_trans.max() > 1:                             
+        img = img_trans / 255.0                          # 保证数据处于0-1之间的浮点数
     return img
 ```
 

Python/Segmentation/Unet.py

@@ -4,17 +4,19 @@ import migraphx
 
 def Preprocessing(pil_img, newW, newH):
     assert newW > 0 and newH > 0, 'Scale is too small' 
-    img_nd = cv2.resize(pil_img, (newW, newH))          # 将图像尺寸修改为256x256
-    img_nd = cv2.cvtColor(img_nd, cv2.COLOR_BGR2RGB)    # BGR转换为RGB
-
+    img_nd = cv2.cvtColor(pil_img, cv2.COLOR_BGR2RGB)    # BGR转换为RGB
+    img_nd = cv2.resize(img_nd, (newW, newH))            # 将图像尺寸修改为256x256
+    
     if len(img_nd.shape) == 2:
         img_nd = np.expand_dims(img_nd, axis=2)
 
-    img_trans = img_nd.transpose((2, 0, 1))             # HWC转换为CHW
-    if img_trans.max() > 1:                             # 保证数据处于0-1之间的浮点数
-        img_trans = img_trans / 255.0  
-    img_trans = np.expand_dims(img_trans, 0)            # CHW扩展为NCHW
-    img = img_trans.astype(np.float32)                  # 转换成浮点型数据
+    img_trans = img_nd.transpose((2, 0, 1))              # HWC转换为CHW
+    img_trans = np.expand_dims(img_trans, 0)             # CHW扩展为NCHW
+    img_trans = np.ascontiguousarray(img_trans)          # 保证内存连续存储
+    img_trans = img_trans.astype(np.float32)             # 转换成浮点型数据
+    if img_trans.max() > 1:                             
+        img = img_trans / 255.0                          # 保证数据处于0-1之间的浮点数
+
     return img
 
 def Sigmoid(x):

README.md

@@ -84,7 +84,7 @@ pip install -r requirements.txt
 python Unet.py
 ```
 
-输出结果为：
+会在当前目录中生成分割图像
 
 <img src="./Doc/Images/Unet_03.jpg" style="zoom:100%;" align=middle>
 
@@ -103,7 +103,7 @@ cd ./build/
 ./MIGraphX_Samples 0
 ```
 
-会在当前目录中生成分割结果图像Result.jpg
+会在当前目录中生成分割图像
 
 <img src="./Doc/Images/Unet_02.jpg" style="zoom:100%;" align=middle>