net_initialize->initialize_net

paddlers/tasks/base.py

@@ -86,7 +86,7 @@ class BaseModel(metaclass=ModelMeta):
         self.quant_config = None
         self.fixed_input_shape = None
 
-    def net_initialize(self,
+    def initialize_net(self,
                        pretrain_weights=None,
                        save_dir='.',
                        resume_checkpoint=None,

paddlers/tasks/change_detector.py

@@ -316,7 +316,7 @@ class BaseChangeDetector(BaseModel):
                         exit=True)
         pretrained_dir = osp.join(save_dir, 'pretrain')
         is_backbone_weights = pretrain_weights == 'IMAGENET'
-        self.net_initialize(
+        self.initialize_net(
             pretrain_weights=pretrain_weights,
             save_dir=pretrained_dir,
             resume_checkpoint=resume_checkpoint,
@@ -607,13 +607,13 @@ class BaseChangeDetector(BaseModel):
             invalid_value (int, optional): Value that marks invalid pixels in output 
                 image. Defaults to 255.
             merge_strategy (str, optional): Strategy to merge overlapping blocks. Choices
-                are {'keep_first', 'keep_last', 'vote'}. 'keep_first' and 'keep_last' 
+                are {'keep_first', 'keep_last', 'accum'}. 'keep_first' and 'keep_last' 
                 means keeping the values of the first and the last block in traversal 
-                order, respectively. 'vote' means applying a simple voting strategy when 
-                there are conflicts in the overlapping pixels. Defaults to 'keep_last'.
+                order, respectively. 'accum' means determining the class of an overlapping 
+                pixel according to accumulated probabilities. Defaults to 'keep_last'.
         """
 
-        slider_predict(self, img_files, save_dir, block_size, overlap,
+        slider_predict(self.predict, img_files, save_dir, block_size, overlap,
                        transforms, invalid_value, merge_strategy)
 
     def preprocess(self, images, transforms, to_tensor=True):

paddlers/tasks/classifier.py

@@ -288,7 +288,7 @@ class BaseClassifier(BaseModel):
                         exit=True)
         pretrained_dir = osp.join(save_dir, 'pretrain')
         is_backbone_weights = False
-        self.net_initialize(
+        self.initialize_net(
             pretrain_weights=pretrain_weights,
             save_dir=pretrained_dir,
             resume_checkpoint=resume_checkpoint,

paddlers/tasks/object_detector.py

@@ -347,7 +347,7 @@ class BaseDetector(BaseModel):
                         "Invalid pretrained weights. Please specify a .pdparams file.",
                         exit=True)
         pretrained_dir = osp.join(save_dir, 'pretrain')
-        self.net_initialize(
+        self.initialize_net(
             pretrain_weights=pretrain_weights,
             save_dir=pretrained_dir,
             resume_checkpoint=resume_checkpoint,

paddlers/tasks/restorer.py

@@ -283,7 +283,7 @@ class BaseRestorer(BaseModel):
                         exit=True)
         pretrained_dir = osp.join(save_dir, 'pretrain')
         is_backbone_weights = pretrain_weights == 'IMAGENET'
-        self.net_initialize(
+        self.initialize_net(
             pretrain_weights=pretrain_weights,
             save_dir=pretrained_dir,
             resume_checkpoint=resume_checkpoint,

paddlers/tasks/segmenter.py

@@ -308,7 +308,7 @@ class BaseSegmenter(BaseModel):
                         exit=True)
         pretrained_dir = osp.join(save_dir, 'pretrain')
         is_backbone_weights = pretrain_weights == 'IMAGENET'
-        self.net_initialize(
+        self.initialize_net(
             pretrain_weights=pretrain_weights,
             save_dir=pretrained_dir,
             resume_checkpoint=resume_checkpoint,
@@ -579,15 +579,13 @@ class BaseSegmenter(BaseModel):
             invalid_value (int, optional): Value that marks invalid pixels in output 
                 image. Defaults to 255.
             merge_strategy (str, optional): Strategy to merge overlapping blocks. Choices
-                are {'keep_first', 'keep_last', 'vote', 'accum'}. 'keep_first' and 
-                'keep_last' means keeping the values of the first and the last block in 
-                traversal order, respectively. 'vote' means applying a simple voting 
-                strategy when there are conflicts in the overlapping pixels. 'accum' 
-                means determining the class of an overlapping pixel according to 
-                accumulated probabilities. Defaults to 'keep_last'.
+                are {'keep_first', 'keep_last', 'accum'}. 'keep_first' and 'keep_last' 
+                means keeping the values of the first and the last block in traversal 
+                order, respectively. 'accum' means determining the class of an overlapping 
+                pixel according to accumulated probabilities. Defaults to 'keep_last'.
         """
 
-        slider_predict(self, img_file, save_dir, block_size, overlap,
+        slider_predict(self.predict, img_file, save_dir, block_size, overlap,
                        transforms, invalid_value, merge_strategy)
 
     def preprocess(self, images, transforms, to_tensor=True):