add unit test for sep_fcn_head: debug 0

mmseg/models/decode_heads/sep_fcn_head.py

@@ -8,7 +8,7 @@ class DepthwiseSeparableFCNHead(FCNHead):
     """Depthwise-Separable Fully Convolutional Network for Semantic
     Segmentation.
 
-    This head is implemented according to Fast-SCNN.
+    This head is implemented according to Fast-SCNN paper.
     Args:
         in_channels(int): Number of output channels of FFM.
         channels(int): Number of middle-stage channels in the decode head.

tests/test_models/test_heads.py

@@ -6,7 +6,8 @@ from mmcv.cnn import ConvModule
 from mmcv.utils.parrots_wrapper import SyncBatchNorm
 
 from mmseg.models.decode_heads import (ANNHead, ASPPHead, CCHead, DAHead,
-                                       DepthwiseSeparableASPPHead, EncHead,
+                                       DepthwiseSeparableASPPHead,
+                                       DepthwiseSeparableFCNHead, EncHead,
                                        FCNHead, GCHead, NLHead, OCRHead,
                                        PSAHead, PSPHead, UPerHead)
 from mmseg.models.decode_heads.decode_head import BaseDecodeHead
@@ -539,3 +540,21 @@ def test_dw_aspp_head():
     assert head.aspp_modules[2].depthwise_conv.dilation == (24, 24)
     outputs = head(inputs)
     assert outputs.shape == (1, head.num_classes, 45, 45)
+
+
+def test_sep_fcn_head():
+    # test sep_fcn_head with concat_input=False
+    head = DepthwiseSeparableFCNHead(
+        in_channels=128,
+        channels=128,
+        concat_input=False,
+        num_classes=19,
+        in_index=-1)
+    x = torch.rand(1, 128, 32, 32)
+    output = head(x)
+    assert output.shape == (1, head.num_classes, 32, 32)
+    assert not head.concat_input
+    from mmseg.ops.separable_conv_module import DepthwiseSeparableConvModule
+    assert isinstance(head.convs[0], DepthwiseSeparableConvModule)
+    assert isinstance(head.convs[1], DepthwiseSeparableConvModule)
+    assert head.conv_seg.kernel_size == (1, 1)