Update pipeline.py

pipeline.py

@@ -35,7 +35,7 @@ class MultiCaReClassifier():
 
     # The outcome dataframe is created.
     self.image_paths = get_image_files(self.image_folder)
-    self.data = pd.DataFrame(columns=[name for name in self.label_dict.keys() if os.path.isdir(os.path.join(models_root, name.replace(':', '_')))])
+    self.data = pd.DataFrame(columns=[name for name in self.label_dict.keys() if os.path.isdir(os.path.join('models', name.replace(':', '_')))])
     self.data['image_path'] = self.image_paths
     self.predict_image_classes()
 
@@ -95,8 +95,8 @@ class MultiCaReClassifier():
 
     '''Method used to identify the corresponding upper model of a given model.'''
 
-    colon_index = search_last_match(model_name, ':')
-    dot_index = search_last_match(model_name, '.')
+    colon_index = self._search_last_match(model_name, ':')
+    dot_index = self._search_last_match(model_name, '.')
     index = max(colon_index, dot_index)
     if index != -1:
       return model_name[:index]
@@ -131,7 +131,8 @@ class MultiCaReClassifier():
     # Models are ran.
     if len(imgs) > 0:
       device = 'cpu'
-      checkpoint_file = os.path.join(self.models_root, model_name.replace(':', '_'), 'model')
+      # checkpoint_file = os.path.join(self.models_root, model_name.replace(':', '_'), 'model')
+      checkpoint_file = os.path.join(model_name.replace(':', '_'), 'model')
       dls = ImageDataLoaders.from_path_func('', imgs, lambda x: '0', item_tfms=Resize((224,224), method='squish'))
       learn = vision_learner(dls, resnet50, n_out=len(labels)).to_fp16()
       learn.load(checkpoint_file, device=device)
@@ -203,4 +204,4 @@ class MultiCaReClassifier():
     for column in column_list:
       if column in label_list:
         label = column
-    return label
+    return label