[Refactor] refactor HRSCDataset

mmrotate/evaluation/metrics/dota_metric.py

@@ -36,9 +36,6 @@ class DOTAMetric(BaseMetric):
         metric (str | list[str]): Metrics to be evaluated. Only support
             'mAP' now. If is list, the first setting in the list will
              be used to evaluate metric.
-        proposal_nums (Sequence[int]): Proposal number used for evaluating
-            recalls, such as recall@100, recall@1000.
-            Defaults to (100, 300, 1000).
         format_only (bool): Format the output results without perform
             evaluation. It is useful when you want to format the result
             to a specific format. Defaults to False.
@@ -50,7 +47,12 @@ class DOTAMetric(BaseMetric):
             patches' results.
         iou_thr (float): IoU threshold of ``nms_rotated`` used in merge
             patches. Defaults to 0.1.
-        version (str): Angle representations. Defaults to 'oc'.
+        angle_version (str): Angle representations. Defaults to 'oc'.
+        eval_mode (str): 'area' or '11points', 'area' means calculating the
+            area under precision-recall curve, '11points' means calculating
+            the average precision of recalls at [0, 0.1, ..., 1].
+            The PASCAL VOC2007 defaults to use '11points', while PASCAL
+            VOC2012 defaults to use 'area'. Defaults to '11points'.
         collect_device (str): Device name used for collecting results from
             different ranks during distributed training. Must be 'cpu' or
             'gpu'. Defaults to 'cpu'.
@@ -66,12 +68,12 @@ def __init__(self,
                  iou_thrs: Union[float, List[float]] = 0.5,
                  scale_ranges: Optional[List[tuple]] = None,
                  metric: Union[str, List[str]] = 'mAP',
-                 proposal_nums: Sequence[int] = (100, 300, 1000),
                  format_only: bool = False,
                  outfile_prefix: Optional[str] = None,
                  merge_patches: bool = False,
                  iou_thr: float = 0.1,
-                 version: str = 'oc',
+                 angle_version: str = 'oc',
+                 eval_mode: str = '11points',
                  collect_device: str = 'cpu',
                  prefix: Optional[str] = None) -> None:
         super().__init__(collect_device=collect_device, prefix=prefix)
@@ -83,12 +85,10 @@ def __init__(self,
         if not isinstance(metric, str):
             assert len(metric) == 1
             metric = metric[0]
-        allowed_metrics = ['recall', 'mAP']
+        allowed_metrics = ['mAP']
         if metric not in allowed_metrics:
-            raise KeyError(
-                f"metric should be one of 'recall', 'mAP', but got {metric}.")
+            raise KeyError(f"metric should be one of 'mAP', but got {metric}.")
         self.metric = metric
-        self.proposal_nums = proposal_nums
 
         self.format_only = format_only
         if self.format_only:
@@ -99,7 +99,10 @@ def __init__(self,
         self.outfile_prefix = outfile_prefix
         self.merge_patches = merge_patches
         self.iou_thr = iou_thr
-        self.version = version
+        self.angle_version = angle_version
+        assert eval_mode in ['area, 11points'], \
+            'Unrecognized mode, only "area" and "11points" are supported'
+        self.use_07_metric = True if eval_mode == '11points' else False
 
     def merge_results(self, results: Sequence[dict],
                       outfile_prefix: str) -> str:
@@ -163,7 +166,7 @@ def merge_results(self, results: Sequence[dict],
             for f, dets in zip(file_objs, dets_per_cls):
                 if dets.size == 0:
                     continue
-                bboxes = obb2poly_np(dets, self.version)
+                bboxes = obb2poly_np(dets, self.angle_version)
                 for bbox in bboxes:
                     txt_element = [img_id, str(bbox[-1])
                                    ] + [f'{p:.2f}' for p in bbox[:-1]]
@@ -300,15 +303,17 @@ def compute_metrics(self, results: list) -> dict:
         if self.metric == 'mAP':
             assert isinstance(self.iou_thrs, list)
             dataset_name = self.dataset_meta['CLASSES']
+            dets = [pred['pred_bbox_scores'] for pred in preds]
 
             mean_aps = []
             for iou_thr in self.iou_thrs:
                 logger.info(f'\n{"-" * 15}iou_thr: {iou_thr}{"-" * 15}')
                 mean_ap, _ = eval_rbbox_map(
-                    preds['pred_bbox_scores'],
+                    dets,
                     gts,
                     scale_ranges=self.scale_ranges,
                     iou_thr=iou_thr,
+                    use_07_metric=self.use_07_metric,
                     dataset=dataset_name,
                     logger=logger)
                 mean_aps.append(mean_ap)

tests/data/hrsc/FullDataSet/Annotations/100000006.xml

@@ -0,0 +1,58 @@
+<HRSC_Image>
+  <Img_ID>100000006</Img_ID>
+  <Place_ID>100000001</Place_ID>
+  <Source_ID>100000001</Source_ID>
+  <Img_NO>100000006</Img_NO>
+  <Img_FileName>100000006</Img_FileName>
+  <Img_FileFmt>bmp</Img_FileFmt>
+  <Img_Date>1900-01-01</Img_Date>
+  <Img_CusType>sealand</Img_CusType>
+  <Img_Des>
+  </Img_Des>
+  <Img_Location>69.040297,33.070036</Img_Location>
+  <Img_SizeWidth>1172</Img_SizeWidth>
+  <Img_SizeHeight>816</Img_SizeHeight>
+  <Img_SizeDepth>3</Img_SizeDepth>
+  <Img_Resolution>1.07</Img_Resolution>
+  <Img_Resolution_Layer>18</Img_Resolution_Layer>
+  <Img_Scale>100</Img_Scale>
+  <Img_SclPxlNum>
+  </Img_SclPxlNum>
+  <segmented>0</segmented>
+  <Img_Havemask>0</Img_Havemask>
+  <Img_MaskFileName>
+  </Img_MaskFileName>
+  <Img_MaskFileFmt>
+  </Img_MaskFileFmt>
+  <Img_MaskType>
+  </Img_MaskType>
+  <Img_SegFileName>
+  </Img_SegFileName>
+  <Img_SegFileFmt>
+  </Img_SegFileFmt>
+  <Img_Rotation>000d</Img_Rotation>
+  <Annotated>1</Annotated>
+  <HRSC_Objects>
+    <HRSC_Object>
+      <Object_ID>100000006</Object_ID>
+      <Class_ID>100000013</Class_ID>
+      <Object_NO>100000006</Object_NO>
+      <truncated>0</truncated>
+      <difficult>0</difficult>
+      <box_xmin>119</box_xmin>
+      <box_ymin>75</box_ymin>
+      <box_xmax>587</box_xmax>
+      <box_ymax>789</box_ymax>
+      <mbox_cx>341.2143</mbox_cx>
+      <mbox_cy>443.3325</mbox_cy>
+      <mbox_w>778.4297</mbox_w>
+      <mbox_h>178.2595</mbox_h>
+      <mbox_ang>-1.122944</mbox_ang>
+      <segmented>0</segmented>
+      <seg_color>
+      </seg_color>
+      <header_x>143</header_x>
+      <header_y>776</header_y>
+    </HRSC_Object>
+  </HRSC_Objects>
+</HRSC_Image>

tests/data/hrsc/demo.txt

@@ -0,0 +1 @@
+100000006

tests/test_datasets/test_dota.py

@@ -8,7 +8,7 @@ class TestDOTADataset(unittest.TestCase):
 
     def test_dota_with_ann_file(self):
         dataset = DOTADataset(
-            data_root='tests/data/',
+            data_root='tests/data/dota/',
             ann_file='labelTxt/',
             data_prefix=dict(img_path='images/'),
             filter_cfg=dict(
@@ -22,7 +22,7 @@ def test_dota_with_ann_file(self):
         self.assertEqual(data_list[0]['img_id'], 'P2805__1024__0___0')
         self.assertEqual(data_list[0]['file_name'], 'P2805__1024__0___0.png')
         self.assertEqual(data_list[0]['img_path'],
-                         'tests/data/images/P2805__1024__0___0.png')
+                         'tests/data/dota/images/P2805__1024__0___0.png')
         self.assertEqual(len(data_list[0]['instances']), 4)
         self.assertEqual(dataset.get_cat_ids(0), [0, 0, 0, 0])
 
@@ -41,6 +41,6 @@ def test_dota_without_ann_file(self):
         self.assertEqual(data_list[0]['img_id'], 'P2805__1024__0___0')
         self.assertEqual(data_list[0]['file_name'], 'P2805__1024__0___0.png')
         self.assertEqual(data_list[0]['img_path'],
-                         'tests/data/images/P2805__1024__0___0.png')
+                         'tests/data/dota/images/P2805__1024__0___0.png')
         self.assertEqual(len(data_list[0]['instances']), 1)
         self.assertEqual(dataset.get_cat_ids(0), [[]])

tests/test_datasets/test_hrsc.py

@@ -0,0 +1,54 @@
+# Copyright (c) OpenMMLab. All rights reserved.
+import unittest
+
+from mmrotate.datasets import HRSCDataset
+
+
+class TestHRSCDataset(unittest.TestCase):
+
+    def test_hrsc(self):
+        dataset = HRSCDataset(
+            data_root='tests/data/hrsc/',
+            ann_file='demo.txt',
+            data_prefix=dict(sub_data_root='FullDataSet/'),
+            filter_cfg=dict(filter_empty_gt=True, min_size=32),
+            pipeline=[])
+        dataset.full_init()
+        self.assertEqual(len(dataset), 1)
+
+        data_list = dataset.load_data_list()
+        self.assertEqual(len(data_list), 1)
+        self.assertEqual(data_list[0]['img_id'], '100000006')
+        self.assertEqual(
+            data_list[0]['img_path'],
+            'tests/data/hrsc/FullDataSet/AllImages/100000006.bmp')
+        self.assertEqual(
+            data_list[0]['xml_path'],
+            'tests/data/hrsc/FullDataSet/Annotations/100000006.xml')
+        self.assertEqual(len(data_list[0]['instances']), 1)
+        self.assertEqual(dataset.get_cat_ids(0), [0])
+        self.assertEqual(dataset._metainfo['CLASSES'], ('ship', ))
+
+    def test_hrsc_classwise(self):
+        dataset = HRSCDataset(
+            data_root='tests/data/hrsc/',
+            ann_file='demo.txt',
+            data_prefix=dict(sub_data_root='FullDataSet/'),
+            classwise=True,
+            filter_cfg=dict(filter_empty_gt=True, min_size=32),
+            pipeline=[])
+        dataset.full_init()
+        self.assertEqual(len(dataset), 1)
+
+        data_list = dataset.load_data_list()
+        self.assertEqual(len(data_list), 1)
+        self.assertEqual(data_list[0]['img_id'], '100000006')
+        self.assertEqual(
+            data_list[0]['img_path'],
+            'tests/data/hrsc/FullDataSet/AllImages/100000006.bmp')
+        self.assertEqual(
+            data_list[0]['xml_path'],
+            'tests/data/hrsc/FullDataSet/Annotations/100000006.xml')
+        self.assertEqual(len(data_list[0]['instances']), 1)
+        self.assertEqual(dataset.get_cat_ids(0), [12])
+        self.assertEqual(len(dataset._metainfo['CLASSES']), 31)