Merge branch 'dev' of https://github.com/open-mmlab/mmrotate into cfg_md

.github/workflows/build.yml

@@ -198,7 +198,7 @@ jobs:
       - name: Build and install
         run: pip install -e .
       - name: Run unittests
-        run: coverage run --branch --source mmrotate -m pytest tests -sv
+        run: coverage run --branch --source mmrotate -m pytest tests
       - name: Generate coverage report
         run: |
           coverage xml

mmrotate/datasets/dota.py

@@ -8,7 +8,7 @@
 import zipfile
 from collections import defaultdict
 from functools import partial
-from multiprocessing import Pool
+from multiprocessing import get_context
 
 import mmcv
 import numpy as np
@@ -380,7 +380,7 @@ def eval_map(det_results,
     area_ranges = ([(rg[0]**2, rg[1]**2) for rg in scale_ranges]
                    if scale_ranges is not None else None)
 
-    pool = Pool(nproc)
+    pool = get_context('spawn').Pool(nproc)
     eval_results = []
     for i in range(num_classes):
         # get gt and det bboxes of this class

mmrotate/models/roi_heads/roi_extractors/rotate_single_level_roi_extractor.py

@@ -98,19 +98,21 @@ def forward(self, feats, rois, roi_scale_factor=None):
         Returns:
             torch.Tensor: Scaled RoI features.
         """
-        out_size = self.roi_layers[0].out_size
+        if isinstance(self.roi_layers[0], ops.RiRoIAlignRotated):
+            out_size = nn.modules.utils._pair(self.roi_layers[0].out_size)
+        else:
+            out_size = self.roi_layers[0].output_size
         num_levels = len(feats)
-        expand_dims = (-1, self.out_channels * out_size * out_size)
+        expand_dims = (-1, self.out_channels * out_size[0] * out_size[1])
         if torch.onnx.is_in_onnx_export():
             # Work around to export mask-rcnn to onnx
             roi_feats = rois[:, :1].clone().detach()
             roi_feats = roi_feats.expand(*expand_dims)
-            roi_feats = roi_feats.reshape(-1, self.out_channels, out_size,
-                                          out_size)
+            roi_feats = roi_feats.reshape(-1, self.out_channels, *out_size)
             roi_feats = roi_feats * 0
         else:
             roi_feats = feats[0].new_zeros(
-                rois.size(0), self.out_channels, out_size, out_size)
+                rois.size(0), self.out_channels, *out_size)
         # TODO: remove this when parrots supports
         if torch.__version__ == 'parrots':
             roi_feats.requires_grad = True