Adding fcn and deeplabv3 directly on mobilenetv3 backbone.

Author: datumbox

torchvision/models/segmentation/segmentation.py

@@ -1,18 +1,22 @@
 from .._utils import IntermediateLayerGetter
 from ..utils import load_state_dict_from_url
+from .. import mobilenet
 from .. import resnet
 from .deeplabv3 import DeepLabHead, DeepLabV3
 from .fcn import FCN, FCNHead
 
 
-__all__ = ['fcn_resnet50', 'fcn_resnet101', 'deeplabv3_resnet50', 'deeplabv3_resnet101']
+__all__ = ['fcn_resnet50', 'fcn_resnet101', 'fcn_mobilenet_v3_large', 'deeplabv3_resnet50', 'deeplabv3_resnet101',
+           'deeplabv3_mobilenet_v3_large']
 
 
 model_urls = {
     'fcn_resnet50_coco': 'https://download.pytorch.org/models/fcn_resnet50_coco-1167a1af.pth',
     'fcn_resnet101_coco': 'https://download.pytorch.org/models/fcn_resnet101_coco-7ecb50ca.pth',
+    'fcn_mobilenet_v3_large_coco': None,
     'deeplabv3_resnet50_coco': 'https://download.pytorch.org/models/deeplabv3_resnet50_coco-cd0a2569.pth',
     'deeplabv3_resnet101_coco': 'https://download.pytorch.org/models/deeplabv3_resnet101_coco-586e9e4e.pth',
+    'deeplabv3_mobilenet_v3_large_coco': None,
 }
 
 
@@ -23,6 +27,15 @@ def _segm_model(name, backbone_name, num_classes, aux, pretrained_backbone=True)
             replace_stride_with_dilation=[False, True, True])
         out_layer = 'layer4'
         aux_layer = 'layer3'
+    elif 'mobilenet' in backbone_name:
+        backbone = mobilenet.__dict__[backbone_name](pretrained=pretrained_backbone).features
+
+        # Gather the indeces of blocks which are strided. These are the locations of C1, ..., Cn-1 blocks.
+        # The first and last blocks are always included because they are the C0 (conv1) and Cn.
+        stage_indices = [0] + [i for i, b in enumerate(backbone) if getattr(b, "is_strided", False)] + [
+            len(backbone) - 1]
+        out_layer = str(stage_indices[-1])
+        aux_layer = str(stage_indices[-2])
     else:
         raise NotImplementedError('backbone {} is not supported as of now'.format(backbone_name))
 
@@ -71,6 +84,8 @@ def fcn_resnet50(pretrained=False, progress=True,
         pretrained (bool): If True, returns a model pre-trained on COCO train2017 which
             contains the same classes as Pascal VOC
         progress (bool): If True, displays a progress bar of the download to stderr
+        num_classes (int): number of output classes of the model (including the background)
+        aux_loss (bool): If True, it uses an auxiliary loss
     """
     return _load_model('fcn', 'resnet50', pretrained, progress, num_classes, aux_loss, **kwargs)
 
@@ -83,10 +98,26 @@ def fcn_resnet101(pretrained=False, progress=True,
         pretrained (bool): If True, returns a model pre-trained on COCO train2017 which
             contains the same classes as Pascal VOC
         progress (bool): If True, displays a progress bar of the download to stderr
+        num_classes (int): number of output classes of the model (including the background)
+        aux_loss (bool): If True, it uses an auxiliary loss
     """
     return _load_model('fcn', 'resnet101', pretrained, progress, num_classes, aux_loss, **kwargs)
 
 
+def fcn_mobilenet_v3_large(pretrained=False, progress=True,
+                           num_classes=21, aux_loss=None, **kwargs):
+    """Constructs a Fully-Convolutional Network model with a MobileNetV3-Large backbone.
+
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on COCO train2017 which
+            contains the same classes as Pascal VOC
+        progress (bool): If True, displays a progress bar of the download to stderr
+        num_classes (int): number of output classes of the model (including the background)
+        aux_loss (bool): If True, it uses an auxiliary loss
+    """
+    return _load_model('fcn', 'mobilenet_v3_large', pretrained, progress, num_classes, aux_loss, **kwargs)
+
+
 def deeplabv3_resnet50(pretrained=False, progress=True,
                        num_classes=21, aux_loss=None, **kwargs):
     """Constructs a DeepLabV3 model with a ResNet-50 backbone.
@@ -95,6 +126,8 @@ def deeplabv3_resnet50(pretrained=False, progress=True,
         pretrained (bool): If True, returns a model pre-trained on COCO train2017 which
             contains the same classes as Pascal VOC
         progress (bool): If True, displays a progress bar of the download to stderr
+        num_classes (int): number of output classes of the model (including the background)
+        aux_loss (bool): If True, it uses an auxiliary loss
     """
     return _load_model('deeplabv3', 'resnet50', pretrained, progress, num_classes, aux_loss, **kwargs)
 
@@ -107,5 +140,21 @@ def deeplabv3_resnet101(pretrained=False, progress=True,
         pretrained (bool): If True, returns a model pre-trained on COCO train2017 which
             contains the same classes as Pascal VOC
         progress (bool): If True, displays a progress bar of the download to stderr
+        num_classes (int): number of output classes of the model (including the background)
+        aux_loss (bool): If True, it uses an auxiliary loss
     """
     return _load_model('deeplabv3', 'resnet101', pretrained, progress, num_classes, aux_loss, **kwargs)
+
+
+def deeplabv3_mobilenet_v3_large(pretrained=False, progress=True,
+                                 num_classes=21, aux_loss=None, **kwargs):
+    """Constructs a DeepLabV3 model with a MobileNetV3-Large backbone.
+
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on COCO train2017 which
+            contains the same classes as Pascal VOC
+        progress (bool): If True, displays a progress bar of the download to stderr
+        num_classes (int): number of output classes of the model (including the background)
+        aux_loss (bool): If True, it uses an auxiliary loss
+    """
+    return _load_model('deeplabv3', 'mobilenet_v3_large', pretrained, progress, num_classes, aux_loss, **kwargs)