update export process

pdseg/models/model_builder.py

@@ -126,24 +126,20 @@ def sigmoid_to_softmax(logit):
 
 def export_preprocess(image):
     """导出模型的预处理流程"""
-    width = cfg.EVAL_CROP_SIZE[0]
-    height = cfg.EVAL_CROP_SIZE[1]
 
     image = fluid.layers.transpose(image, [0, 3, 1, 2])
     origin_shape = fluid.layers.shape(image)[-2:]
 
     # 不同AUG_METHOD方法的resize
     if cfg.AUG.AUG_METHOD == 'unpadding':
-        h = cfg.AUG.FIX_RESIZE_SIZE[1]
-        w = cfg.AUG.FIX_RESIZE_SIZE[0]
+        h_fix = cfg.AUG.FIX_RESIZE_SIZE[1]
+        w_fix = cfg.AUG.FIX_RESIZE_SIZE[0]
         image = fluid.layers.resize_bilinear(
             image,
-            out_shape=[h, w],
+            out_shape=[h_fix, w_fix],
             align_corners=False,
             align_mode=0)
-    if cfg.AUG.AUG_METHOD == 'stepscaling':
-        pass
-    if cfg.AUG.AUG_METHOD == 'rangescaling':
+    elif cfg.AUG.AUG_METHOD == 'rangescaling':
         size = cfg.AUG.INF_RESIZE_VALUE
         value = fluid.layers.reduce_max(origin_shape)
         scale = float(size) / value.astype('float32')
@@ -153,7 +149,9 @@ def export_preprocess(image):
     # 存储resize后图像shape
     valid_shape = fluid.layers.shape(image)[-2:]
 
-    # padding 到eval_crop_size大小
+    # padding到eval_crop_size大小
+    width = cfg.EVAL_CROP_SIZE[0]
+    height = cfg.EVAL_CROP_SIZE[1]
     pad_target = fluid.layers.assign(
         np.array([height, width]).astype('float32'))
     up = fluid.layers.assign(np.array([0]).astype('float32'))
@@ -171,7 +169,7 @@ def export_preprocess(image):
     std = np.array(cfg.STD).reshape(1, len(cfg.STD), 1, 1)
     std = fluid.layers.assign(std.astype('float32'))
     image = (image / 255 - mean) / std
-    # 很有必要，使后面的网络能通过image.shape获取特征图的shape
+    # 使后面的网络能通过类似image.shape获取特征图的shape
     image = fluid.layers.reshape(
         image, shape=[-1, cfg.DATASET.DATA_DIM, height, width])
     return image, valid_shape, origin_shape

pdseg/utils/collect.py

@@ -122,6 +122,12 @@ class SegConfig(dict):
                 len(self.MODEL.MULTI_LOSS_WEIGHT) != 3:
             self.MODEL.MULTI_LOSS_WEIGHT = [1.0, 0.4, 0.16]
 
+        if self.AUG.AUG_METHOD not in ['unpadding', 'stepscaling', 'rangescaling']:
+            raise ValueError(
+                'AUG.AUG_METHOD config error, only support `unpadding`, `unpadding` and `rangescaling`'
+            )
+
+
     def update_from_list(self, config_list):
         if len(config_list) % 2 != 0:
             raise ValueError(