Internal PR, enable the segement_only flag

demo/image_demo.py

@@ -21,6 +21,10 @@ def main():
         help='Opacity of painted segmentation map. In (0, 1] range.')
     parser.add_argument(
         '--title', default='result', help='The image identifier.')
+    parser.add_argument(
+        '--segementation-only',
+        action='store_true',
+        help='Only show the segmentation map.')
     args = parser.parse_args()
 
     # build the model from a config file and a checkpoint file
@@ -38,7 +42,8 @@ def main():
         opacity=args.opacity,
         draw_gt=False,
         show=False if args.out_file is not None else True,
-        out_file=args.out_file)
+        out_file=args.out_file,
+        segmentation_only=args.segementation_only)
 
 
 if __name__ == '__main__':

docs/en/user_guides/visualization.md

@@ -137,6 +137,8 @@ gt_sem_seg = PixelData(**gt_sem_seg_data)
 data_sample = SegDataSample()
 data_sample.gt_sem_seg = gt_sem_seg
 
+# If you want to show the segement data only (without lable text and background)
+# you can use segement_only = True.
 seg_local_visualizer = SegLocalVisualizer(
     vis_backends=[dict(type='LocalVisBackend')],
     save_dir=save_dir)

mmseg/apis/inference.py

@@ -128,6 +128,7 @@ def show_result_pyplot(model: BaseSegmentor,
                        wait_time: float = 0,
                        show: bool = True,
                        withLabels: Optional[bool] = True,
+                       segmentation_only: bool = False,
                        save_dir=None,
                        out_file=None):
     """Visualize the segmentation results on the image.
@@ -167,10 +168,18 @@ def show_result_pyplot(model: BaseSegmentor,
     if save_dir is not None:
         mkdir_or_exist(save_dir)
     # init visualizer
+    if withLabels and segmentation_only:
+        # Issue a warning if withLabels is True and segmentation_only is True
+        warnings.simplefilter('once')
+        warnings.warn(
+            'withLabels is True and segmentation_only is True, '
+            'withLabels will be set to False')
+        withLabels = False
     visualizer = SegLocalVisualizer(
         vis_backends=[dict(type='LocalVisBackend')],
         save_dir=save_dir,
-        alpha=opacity)
+        alpha=opacity,
+        segement_only=segmentation_only)
     visualizer.dataset_meta = dict(
         classes=model.dataset_meta['classes'],
         palette=model.dataset_meta['palette'])

mmseg/visualization/local_visualizer.py

@@ -39,6 +39,11 @@ class SegLocalVisualizer(Visualizer):
             Defaults to None.
         alpha (int, float): The transparency of segmentation mask.
                 Defaults to 0.8.
+        segement_only (bool): Only show the segmentation map, by setting this flag
+            to True, only the segmentation map will be shown, the background image
+            and the labels will not be shown. Defaults to False.
+            Warning: If withLabels is True and segmentation_only is True,
+            withLabels will be set to False. i.e. this flag will override withLabels.
 
     Examples:
         >>> import numpy as np
@@ -73,10 +78,14 @@ def __init__(self,
                  palette: Optional[List] = None,
                  dataset_name: Optional[str] = None,
                  alpha: float = 0.8,
+                 segement_only: bool = False,
                  **kwargs):
         super().__init__(name, image, vis_backends, save_dir, **kwargs)
         self.alpha: float = alpha
+        if segement_only:
+            self.alpha = 1
         self.set_dataset_meta(palette, classes, dataset_name)
+        self.segement_only = segement_only
 
     def _get_center_loc(self, mask: np.ndarray) -> np.ndarray:
         """Get semantic seg center coordinate.
@@ -139,7 +148,7 @@ def _draw_sem_seg(self,
         for label, color in zip(labels, colors):
             mask[sem_seg[0] == label, :] = color
 
-        if withLabels:
+        if withLabels and not self.segement_only:
             font = cv2.FONT_HERSHEY_SIMPLEX
             # (0,1] to change the size of the text relative to the image
             scale = 0.05