Update mask_adapter/sam_maskadapter.py

mask_adapter/sam_maskadapter.py

@@ -113,7 +113,7 @@ class SAMVisualizationDemo(object):
         
         return clip_vis_dense
     
-    def run_on_image(self, ori_image, class_names):
+    def run_on_image(self, ori_image, class_names, text_features):
         height, width, _ = ori_image.shape
         if width > height:
             new_width = 896
@@ -140,15 +140,15 @@ class SAMVisualizationDemo(object):
 
         image = image.unsqueeze(0)
         
-        if len(class_names) == 1:
-            class_names.append('others')
-        txts = [f'a photo of {cls_name}' for cls_name in class_names]
-        text = open_clip.tokenize(txts)
+        # if len(class_names) == 1:
+        #     class_names.append('others')
+        # txts = [f'a photo of {cls_name}' for cls_name in class_names]
+        # text = open_clip.tokenize(txts)
 
 
         with torch.no_grad():
-            text_features = self.clip_model.encode_text(text)
-            text_features /= text_features.norm(dim=-1, keepdim=True)
+            # text_features = self.clip_model.encode_text(text)
+            # text_features /= text_features.norm(dim=-1, keepdim=True)
             
             features = self.extract_features_convnext(image.float())