implement wide resnet

Author: LyndonKong

python/paddle/tests/test_vision_models.py

@@ -71,6 +71,15 @@ def test_resnet101(self):
     def test_resnet152(self):
         self.models_infer('resnet152')
 
+    def test_wide_resnet50(self):
+        self.models_infer('wide_resnet50')
+
+    def test_wide_resnet101(self):
+        self.models_infer('wide_resnet101')
+
+    def test_wide_resnet101_pretrained(self):
+        self.models_infer('wide_resnet101', pretrained=False)
+
     def test_vgg16_num_classes(self):
         vgg16 = models.__dict__['vgg16'](pretrained=False, num_classes=10)
 

python/paddle/vision/__init__.py

@@ -34,6 +34,9 @@
 from .models import resnet50  # noqa: F401
 from .models import resnet101  # noqa: F401
 from .models import resnet152  # noqa: F401
+from .models import WideResNet  # noqa: F401
+from .models import wide_resnet50  # noqa: F401
+from .models import wide_resnet101  # noqa: F401
 from .models import MobileNetV1  # noqa: F401
 from .models import mobilenet_v1  # noqa: F401
 from .models import MobileNetV2  # noqa: F401

python/paddle/vision/models/__init__.py

@@ -18,6 +18,9 @@
 from .resnet import resnet50  # noqa: F401
 from .resnet import resnet101  # noqa: F401
 from .resnet import resnet152  # noqa: F401
+from .wideresnet import WideResNet  # noqa: F401
+from .wideresnet import wide_resnet50  # noqa: F401
+from .wideresnet import wide_resnet101  # noqa: F401
 from .mobilenetv1 import MobileNetV1  # noqa: F401
 from .mobilenetv1 import mobilenet_v1  # noqa: F401
 from .mobilenetv2 import MobileNetV2  # noqa: F401

python/paddle/vision/models/resnet.py

@@ -155,6 +155,7 @@ class ResNet(nn.Layer):
         depth (int): layers of resnet, default: 50.
         num_classes (int): output dim of last fc layer. If num_classes <=0, last fc layer 
                             will not be defined. Default: 1000.
+        width_per_group (int): 
         with_pool (bool): use pool before the last fc layer or not. Default: True.
 
     Examples:
@@ -169,7 +170,12 @@ class ResNet(nn.Layer):
 
     """
 
-    def __init__(self, block, depth, num_classes=1000, with_pool=True):
+    def __init__(self,
+                 block,
+                 depth,
+                 num_classes=1000,
+                 width_per_group=64,
+                 with_pool=True):
         super(ResNet, self).__init__()
         layer_cfg = {
             18: [2, 2, 2, 2],
@@ -180,6 +186,7 @@ def __init__(self, block, depth, num_classes=1000, with_pool=True):
         }
         layers = layer_cfg[depth]
         self.num_classes = num_classes
+        self.base_width = width_per_group
         self.with_pool = with_pool
         self._norm_layer = nn.BatchNorm2D
 
@@ -225,11 +232,16 @@ def _make_layer(self, block, planes, blocks, stride=1, dilate=False):
 
         layers = []
         layers.append(
-            block(self.inplanes, planes, stride, downsample, 1, 64,
+            block(self.inplanes, planes, stride, downsample, 1, self.base_width,
                   previous_dilation, norm_layer))
         self.inplanes = planes * block.expansion
         for _ in range(1, blocks):
-            layers.append(block(self.inplanes, planes, norm_layer=norm_layer))
+            layers.append(
+                block(
+                    self.inplanes,
+                    planes,
+                    base_width=self.base_width,
+                    norm_layer=norm_layer))
 
         return nn.Sequential(*layers)
 

python/paddle/vision/models/wideresnet.py

@@ -0,0 +1,121 @@
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import division
+from __future__ import print_function
+
+import paddle
+import paddle.nn as nn
+
+from paddle.utils.download import get_weights_path_from_url
+from paddle.vision.models.resnet import BottleneckBlock, ResNet
+
+__all__ = []
+
+model_urls = {'wide_resnet50': ('', ''), 'wide_resnet101': ('', '')}
+
+
+class WideResNet(nn.Layer):
+    """Wide ResNet model from
+    `"Wide Residual Networks" <https://arxiv.org/pdf/1605.07146.pdf>`_.
+
+    The model is the same as ResNet except for the bottleneck number of channels
+    which is twice larger in every block. The number of channels in outer 1x1
+    convolutions is the same.
+
+    Args:
+        Block (BasicBlock|BottleneckBlock): block module of model.
+        depth (int): layers of resnet, default: 50.
+        num_classes (int): output dim of last fc layer. If num_classes <=0, last fc layer 
+                            will not be defined. Default: 1000.
+        width_per_group (int): channel nums of each group
+        with_pool (bool): use pool before the last fc layer or not. Default: True.
+
+    Examples:
+        .. code-block:: python
+
+            from paddle.vision.models import WideResNet
+
+            wide_resnet50 = WideResNet(50)
+
+            wide_resnet101 = WideResNet(101)
+
+    """
+
+    def __init__(self,
+                 depth,
+                 num_classes=1000,
+                 width_per_group=64,
+                 with_pool=True):
+        super(WideResNet, self).__init__()
+        self.layers = ResNet(BottleneckBlock, depth, num_classes,
+                             width_per_group * 2, with_pool)
+
+    def forward(self, x):
+        return self.layers.forward(x)
+
+
+def _wide_resnet(arch, depth, pretrained, **kwargs):
+    model = WideResNet(depth, **kwargs)
+    if pretrained:
+        assert arch in model_urls, "{} model do not have a pretrained model now, you should set pretrained=False".format(
+            arch)
+        weight_path = get_weights_path_from_url(model_urls[arch][0],
+                                                model_urls[arch][1])
+
+        param = paddle.load(weight_path)
+        model.set_dict(param)
+
+    return model
+
+
+def wide_resnet50(pretrained=False, **kwargs):
+    """Wide ResNet 50-layer model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+
+    Examples:
+        .. code-block:: python
+
+            from paddle.vision.models import wide_resnet50
+
+            # build model
+            model = wide_resnet50()
+
+            # build model and load imagenet pretrained weight
+            # model = wide_resnet50(pretrained=True)
+    """
+    kwargs['width_per_group'] = 64 * 2
+    return _wide_resnet('wide_resnet50', 50, pretrained, **kwargs)
+
+
+def wide_resnet101(pretrained=False, **kwargs):
+    """Wide ResNet 101-layer model
+    
+    Args:
+        pretrained (bool): If True, returns a model pre-trained on ImageNet
+
+    Examples:
+        .. code-block:: python
+
+            from paddle.vision.models import wide_resnet101
+
+            # build model
+            model = wide_resnet101()
+
+            # build model and load imagenet pretrained weight
+            # model = wide_resnet101(pretrained=True)
+    """
+    return _wide_resnet('wide_resnet101', 101, pretrained, **kwargs)