[Fix]Add label_map and reduce_zero_label in metainfo of dataset and deprecate reduce_zero_label in load annotation

mmseg/datasets/custom.py

@@ -122,6 +122,10 @@ class CustomDataset(BaseDataset):
         # Get label map for custom classes
         new_classes = self._metainfo.get('classes', None)
         self.label_map = self.get_label_map(new_classes)
+        self._metainfo.update(
+            dict(
+                label_map=self.label_map,
+                reduce_zero_label=self.reduce_zero_label))
 
         # Update palette based on label map or generate palette
         # if it is not defined
@@ -240,7 +244,8 @@ class CustomDataset(BaseDataset):
                     seg_map = img_name + self.seg_map_suffix
                     data_info['seg_map_path'] = osp.join(ann_dir, seg_map)
                 data_info['label_map'] = self.label_map
-                data_info['seg_field'] = []
+                data_info['reduce_zero_label'] = self.reduce_zero_label
+                data_info['seg_fields'] = []
                 data_list.append(data_info)
         else:
             img_dir = self.data_prefix['img_path']
@@ -254,7 +259,8 @@ class CustomDataset(BaseDataset):
                     seg_map = img.replace(self.img_suffix, self.seg_map_suffix)
                     data_info['seg_map_path'] = osp.join(ann_dir, seg_map)
                 data_info['label_map'] = self.label_map
-                data_info['seg_field'] = []
+                data_info['reduce_zero_label'] = self.reduce_zero_label
+                data_info['seg_fields'] = []
                 data_list.append(data_info)
             data_list = sorted(data_list, key=lambda x: x['img_path'])
         return data_list

mmseg/datasets/pipelines/loading.py

@@ -1,4 +1,6 @@
 # Copyright (c) OpenMMLab. All rights reserved.
+import warnings
+
 import mmcv
 import numpy as np
 from mmcv.transforms import LoadAnnotations as MMCV_LoadAnnotations
@@ -40,9 +42,9 @@ class LoadAnnotations(MMCV_LoadAnnotations):
     - gt_seg_map (np.uint8)
 
     Args:
-        reduce_zero_label (bool): Whether reduce all label value by 1.
+        reduce_zero_label (bool, optional): Whether reduce all label value
-            Usually used for datasets where 0 is background label.
+            by 1. Usually used for datasets where 0 is background label.
-            Defaults to False.
+            Defaults to None.
         imdecode_backend (str): The image decoding backend type. The backend
             argument for :func:``mmcv.imfrombytes``.
             See :fun:``mmcv.imfrombytes`` for details.
@@ -54,7 +56,7 @@ class LoadAnnotations(MMCV_LoadAnnotations):
 
     def __init__(
         self,
-        reduce_zero_label=False,
+        reduce_zero_label=None,
         file_client_args=dict(backend='disk'),
         imdecode_backend='pillow',
     ) -> None:
@@ -66,6 +68,11 @@ class LoadAnnotations(MMCV_LoadAnnotations):
             imdecode_backend=imdecode_backend,
             file_client_args=file_client_args)
         self.reduce_zero_label = reduce_zero_label
+        if self.reduce_zero_label is not None:
+            warnings.warn('`reduce_zero_label` will be deprecated, '
+                          'if you would like to ignore the zero label, please '
+                          'set `reduce_zero_label=True` when dataset '
+                          'initialized')
         self.file_client_args = file_client_args.copy()
         self.imdecode_backend = imdecode_backend
 
@@ -93,6 +100,12 @@ class LoadAnnotations(MMCV_LoadAnnotations):
             for old_id, new_id in results['label_map'].items():
                 gt_semantic_seg[gt_semantic_seg_copy == old_id] = new_id
         # reduce zero_label
+        if self.reduce_zero_label is None:
+            self.reduce_zero_label = results['reduce_zero_label']
+        assert self.reduce_zero_label == results['reduce_zero_label'], \
+            'Initialize dataset with `reduce_zero_label` as ' \
+            f'{results["reduce_zero_label"]} but when load annotation ' \
+            f'the `reduce_zero_label` is {self.reduce_zero_label}'
         if self.reduce_zero_label:
             # avoid using underflow conversion
             gt_semantic_seg[gt_semantic_seg == 0] = 255

tests/test_datasets/test_loading.py

@@ -47,13 +47,14 @@ class TestLoading(object):
 
     def test_load_seg(self):
         seg_path = osp.join(self.data_prefix, 'seg.png')
-        results = dict(seg_map_path=seg_path, seg_fields=[])
+        results = dict(
+            seg_map_path=seg_path, reduce_zero_label=True, seg_fields=[])
         transform = LoadAnnotations()
         results = transform(copy.deepcopy(results))
         assert results['gt_seg_map'].shape == (288, 512)
         assert results['gt_seg_map'].dtype == np.uint8
         assert repr(transform) == transform.__class__.__name__ + \
-            "(reduce_zero_label=False,imdecode_backend='pillow')" + \
+            "(reduce_zero_label=True,imdecode_backend='pillow')" + \
             "file_client_args={'backend': 'disk'})"
 
         # reduce_zero_label
@@ -89,6 +90,7 @@ class TestLoading(object):
                 3: 1,
                 4: 0
             },
+            reduce_zero_label=False,
             seg_fields=[])
 
         load_imgs = LoadImageFromFile()
@@ -118,6 +120,7 @@ class TestLoading(object):
                 3: 2,
                 4: 1
             },
+            reduce_zero_label=False,
             seg_fields=[])
 
         load_imgs = LoadImageFromFile()
@@ -138,7 +141,11 @@ class TestLoading(object):
         np.testing.assert_array_equal(gt_array, true_mask)
 
         # test no custom classes
-        results = dict(img_path=img_path, seg_map_path=gt_path, seg_fields=[])
+        results = dict(
+            img_path=img_path,
+            seg_map_path=gt_path,
+            reduce_zero_label=False,
+            seg_fields=[])
 
         load_imgs = LoadImageFromFile()
         results = load_imgs(copy.deepcopy(results))

tests/test_datasets/test_pipelines/test_transforms.py

@@ -49,7 +49,7 @@ def test_photo_metric_distortion():
     results['pad_shape'] = img.shape
     results['scale_factor'] = 1.0
 
-    pipeline = PhotoMetricDistortion()
+    pipeline = PhotoMetricDistortion(saturation_range=(1., 1.))
     results = pipeline(results)
 
     assert not ((results['img'] == img).all())