Add support for custom classes (open-mmlab#71)

* Support for custom classes

* Fix test

* Fix pre-commit

* Add pipeline logic for custom classes

* Fix minor issues, fix test

* Fix issues from PR review

* Fix tests

* Remove palette as str

* Rename old_to_new_ids to label_map

* Test for load_anns

* Remove get_palette function

* fixed temp

* Add subset of palette, remove palette as arg

* minor update

Co-authored-by: Jiarui XU <[email protected]>

Author: igonro

mmseg/datasets/custom.py

@@ -58,6 +58,8 @@ class CustomDataset(Dataset):
         ignore_index (int): The label index to be ignored. Default: 255
         reduce_zero_label (bool): Whether to mark label zero as ignored.
             Default: False
+        classes (str | Sequence[str], optional): Specify classes to load.
+            If is None, ``cls.CLASSES`` will be used. Default: None.
     """
 
     CLASSES = None
@@ -74,7 +76,8 @@ def __init__(self,
                  data_root=None,
                  test_mode=False,
                  ignore_index=255,
-                 reduce_zero_label=False):
+                 reduce_zero_label=False,
+                 classes=None):
         self.pipeline = Compose(pipeline)
         self.img_dir = img_dir
         self.img_suffix = img_suffix
@@ -85,6 +88,8 @@ def __init__(self,
         self.test_mode = test_mode
         self.ignore_index = ignore_index
         self.reduce_zero_label = reduce_zero_label
+        self.label_map = None
+        self.CLASSES, self.PALETTE = self.get_classes_and_palette(classes)
 
         # join paths if data_root is specified
         if self.data_root is not None:
@@ -160,6 +165,8 @@ def get_ann_info(self, idx):
     def pre_pipeline(self, results):
         """Prepare results dict for pipeline."""
         results['seg_fields'] = []
+        if self.custom_classes:
+            results['label_map'] = self.label_map
 
     def __getitem__(self, idx):
         """Get training/test data after pipeline.
@@ -220,6 +227,10 @@ def get_gt_seg_maps(self):
         for img_info in self.img_infos:
             gt_seg_map = mmcv.imread(
                 img_info['ann']['seg_map'], flag='unchanged', backend='pillow')
+            # modify if custom classes
+            if self.label_map is not None:
+                for old_id, new_id in self.label_map.items():
+                    gt_seg_map[gt_seg_map == old_id] = new_id
             if self.reduce_zero_label:
                 # avoid using underflow conversion
                 gt_seg_map[gt_seg_map == 0] = 255
@@ -230,6 +241,63 @@ def get_gt_seg_maps(self):
 
         return gt_seg_maps
 
+    def get_classes_and_palette(self, classes=None):
+        """Get class names of current dataset.
+
+        Args:
+            classes (Sequence[str] | str | None): If classes is None, use
+                default CLASSES defined by builtin dataset. If classes is a
+                string, take it as a file name. The file contains the name of
+                classes where each line contains one class name. If classes is
+                a tuple or list, override the CLASSES defined by the dataset.
+        """
+        if classes is None:
+            self.custom_classes = False
+            return self.CLASSES, self.PALETTE
+
+        self.custom_classes = True
+        if isinstance(classes, str):
+            # take it as a file path
+            class_names = mmcv.list_from_file(classes)
+        elif isinstance(classes, (tuple, list)):
+            class_names = classes
+        else:
+            raise ValueError(f'Unsupported type {type(classes)} of classes.')
+
+        if self.CLASSES:
+            if not set(classes).issubset(self.CLASSES):
+                raise ValueError('classes is not a subset of CLASSES.')
+
+            # dictionary, its keys are the old label ids and its values
+            # are the new label ids.
+            # used for changing pixel labels in load_annotations.
+            self.label_map = {}
+            for i, c in enumerate(self.CLASSES):
+                if c not in class_names:
+                    self.label_map[i] = -1
+                else:
+                    self.label_map[i] = classes.index(c)
+
+        palette = self.get_palette_for_custom_classes()
+
+        return class_names, palette
+
+    def get_palette_for_custom_classes(self):
+
+        if self.label_map is not None:
+            # return subset of palette
+            palette = []
+            for old_id, new_id in sorted(
+                    self.label_map.items(), key=lambda x: x[1]):
+                if new_id != -1:
+                    palette.append(self.PALETTE[old_id])
+            palette = type(self.PALETTE)(palette)
+
+        else:
+            palette = self.PALETTE
+
+        return palette
+
     def evaluate(self, results, metric='mIoU', logger=None, **kwargs):
         """Evaluate the dataset.
 

mmseg/datasets/pipelines/loading.py

@@ -132,6 +132,10 @@ def __call__(self, results):
         gt_semantic_seg = mmcv.imfrombytes(
             img_bytes, flag='unchanged',
             backend=self.imdecode_backend).squeeze().astype(np.uint8)
+        # modify if custom classes
+        if results.get('label_map', None) is not None:
+            for old_id, new_id in results['label_map'].items():
+                gt_semantic_seg[gt_semantic_seg == old_id] = new_id
         # reduce zero_label
         if self.reduce_zero_label:
             # avoid using underflow conversion

tests/test_data/test_dataset.py

@@ -5,8 +5,9 @@
 import pytest
 
 from mmseg.core.evaluation import get_classes, get_palette
-from mmseg.datasets import (ADE20KDataset, CityscapesDataset, ConcatDataset,
-                            CustomDataset, PascalVOCDataset, RepeatDataset)
+from mmseg.datasets import (DATASETS, ADE20KDataset, CityscapesDataset,
+                            ConcatDataset, CustomDataset, PascalVOCDataset,
+                            RepeatDataset)
 
 
 def test_classes():
@@ -171,3 +172,62 @@ def test_custom_dataset():
     assert 'mIoU' in eval_results
     assert 'mAcc' in eval_results
     assert 'aAcc' in eval_results
+
+
+@patch('mmseg.datasets.CustomDataset.load_annotations', MagicMock)
+@patch('mmseg.datasets.CustomDataset.__getitem__',
+       MagicMock(side_effect=lambda idx: idx))
+@pytest.mark.parametrize('dataset, classes', [
+    ('ADE20KDataset', ('wall', 'building')),
+    ('CityscapesDataset', ('road', 'sidewalk')),
+    ('CustomDataset', ('bus', 'car')),
+    ('PascalVOCDataset', ('aeroplane', 'bicycle')),
+])
+def test_custom_classes_override_default(dataset, classes):
+
+    dataset_class = DATASETS.get(dataset)
+
+    original_classes = dataset_class.CLASSES
+
+    # Test setting classes as a tuple
+    custom_dataset = dataset_class(
+        pipeline=[],
+        img_dir=MagicMock(),
+        split=MagicMock(),
+        classes=classes,
+        test_mode=True)
+
+    assert custom_dataset.CLASSES != original_classes
+    assert custom_dataset.CLASSES == classes
+
+    # Test setting classes as a list
+    custom_dataset = dataset_class(
+        pipeline=[],
+        img_dir=MagicMock(),
+        split=MagicMock(),
+        classes=list(classes),
+        test_mode=True)
+
+    assert custom_dataset.CLASSES != original_classes
+    assert custom_dataset.CLASSES == list(classes)
+
+    # Test overriding not a subset
+    custom_dataset = dataset_class(
+        pipeline=[],
+        img_dir=MagicMock(),
+        split=MagicMock(),
+        classes=[classes[0]],
+        test_mode=True)
+
+    assert custom_dataset.CLASSES != original_classes
+    assert custom_dataset.CLASSES == [classes[0]]
+
+    # Test default behavior
+    custom_dataset = dataset_class(
+        pipeline=[],
+        img_dir=MagicMock(),
+        split=MagicMock(),
+        classes=None,
+        test_mode=True)
+
+    assert custom_dataset.CLASSES == original_classes

tests/test_data/test_loading.py

@@ -1,6 +1,8 @@
 import copy
 import os.path as osp
+import tempfile
 
+import mmcv
 import numpy as np
 
 from mmseg.datasets.pipelines import LoadAnnotations, LoadImageFromFile
@@ -98,3 +100,99 @@ def test_load_seg(self):
         # this image is saved by PIL
         assert results['gt_semantic_seg'].shape == (288, 512)
         assert results['gt_semantic_seg'].dtype == np.uint8
+
+    def test_load_seg_custom_classes(self):
+
+        test_img = np.random.rand(10, 10)
+        test_gt = np.zeros_like(test_img)
+        test_gt[2:4, 2:4] = 1
+        test_gt[2:4, 6:8] = 2
+        test_gt[6:8, 2:4] = 3
+        test_gt[6:8, 6:8] = 4
+
+        tmp_dir = tempfile.TemporaryDirectory()
+        img_path = osp.join(tmp_dir.name, 'img.jpg')
+        gt_path = osp.join(tmp_dir.name, 'gt.png')
+
+        mmcv.imwrite(test_img, img_path)
+        mmcv.imwrite(test_gt, gt_path)
+
+        # test only train with label with id 3
+        results = dict(
+            img_info=dict(filename=img_path),
+            ann_info=dict(seg_map=gt_path),
+            label_map={
+                0: 0,
+                1: 0,
+                2: 0,
+                3: 1,
+                4: 0
+            },
+            seg_fields=[])
+
+        load_imgs = LoadImageFromFile()
+        results = load_imgs(copy.deepcopy(results))
+
+        load_anns = LoadAnnotations()
+        results = load_anns(copy.deepcopy(results))
+
+        gt_array = results['gt_semantic_seg']
+
+        true_mask = np.zeros_like(gt_array)
+        true_mask[6:8, 2:4] = 1
+
+        assert results['seg_fields'] == ['gt_semantic_seg']
+        assert gt_array.shape == (10, 10)
+        assert gt_array.dtype == np.uint8
+        np.testing.assert_array_equal(gt_array, true_mask)
+
+        # test only train with label with id 4 and 3
+        results = dict(
+            img_info=dict(filename=img_path),
+            ann_info=dict(seg_map=gt_path),
+            label_map={
+                0: 0,
+                1: 0,
+                2: 0,
+                3: 2,
+                4: 1
+            },
+            seg_fields=[])
+
+        load_imgs = LoadImageFromFile()
+        results = load_imgs(copy.deepcopy(results))
+
+        load_anns = LoadAnnotations()
+        results = load_anns(copy.deepcopy(results))
+
+        gt_array = results['gt_semantic_seg']
+
+        true_mask = np.zeros_like(gt_array)
+        true_mask[6:8, 2:4] = 2
+        true_mask[6:8, 6:8] = 1
+
+        assert results['seg_fields'] == ['gt_semantic_seg']
+        assert gt_array.shape == (10, 10)
+        assert gt_array.dtype == np.uint8
+        np.testing.assert_array_equal(gt_array, true_mask)
+
+        # test no custom classes
+        results = dict(
+            img_info=dict(filename=img_path),
+            ann_info=dict(seg_map=gt_path),
+            seg_fields=[])
+
+        load_imgs = LoadImageFromFile()
+        results = load_imgs(copy.deepcopy(results))
+
+        load_anns = LoadAnnotations()
+        results = load_anns(copy.deepcopy(results))
+
+        gt_array = results['gt_semantic_seg']
+
+        assert results['seg_fields'] == ['gt_semantic_seg']
+        assert gt_array.shape == (10, 10)
+        assert gt_array.dtype == np.uint8
+        np.testing.assert_array_equal(gt_array, test_gt)
+
+        tmp_dir.cleanup()