Delete 'model/backbone/MobileNetV2.py'

model/backbone/MobileNetV2.py

@@ -1,103 +0,0 @@
-"""
-MindSpore implementation of `MobileNetV2`.
-Refer to "MobileNetV2: Inverted Residuals and Linear Bottlenecks"
-"""
-import mindspore as ms
-from mindspore import nn
-
-
-class InvertedBlock(nn.Cell):
-    """ Inverted Block """
-    def __init__(self, in_chans, out_chans, expand_ratio, stride):
-        super().__init__()
-        self.stride = stride
-        assert stride in [1, 2]
-
-        hidden_dim = in_chans * expand_ratio
-        self.use_res_connect = self.stride == 1 and in_chans == out_chans
-
-        layers = []
-        if expand_ratio != 1:
-            layers.append(nn.SequentialCell(
-                nn.Conv2d(in_chans, hidden_dim, kernel_size=3,
-                          pad_mode='pad', padding=1, stride=stride, group=in_chans, has_bias=False),
-                nn.BatchNorm2d(hidden_dim),
-                nn.ReLU6()
-            ))
-        layers.append(nn.SequentialCell(
-            nn.Conv2d(hidden_dim, hidden_dim, kernel_size=3,
-                      pad_mode='pad', padding=1, stride=stride, group=in_chans, has_bias=False),
-            nn.BatchNorm2d(hidden_dim),
-            nn.ReLU6()
-        ))
-        layers.append(nn.SequentialCell(
-            nn.Conv2d(hidden_dim, out_chans, kernel_size=1, stride=1, has_bias=False),
-            nn.BatchNorm2d(out_chans),
-            nn.ReLU6()
-        ))
-
-        self.layers = nn.SequentialCell(*layers)
-
-    def construct(self, x):
-        if self.use_res_connect:
-            return x + self.layers(x)
-
-        return self.layers(x)
-
-
-class MobileNetV2(nn.Cell):
-    """ MobileNet V2 """
-    def __init__(self, in_chans=3, num_classes=1000):
-        super().__init__()
-        self.configs = [
-            # t, c, n, s
-            [1, 16, 1, 1],
-            [6, 24, 2, 2],
-            [6, 32, 3, 2],
-            [6, 64, 4, 2],
-            [6, 96, 3, 1],
-            [6, 160, 3, 2],
-            [6, 320, 1, 1]
-        ]
-
-        self.stem_conv = nn.SequentialCell(
-            nn.Conv2d(in_chans, 32, kernel_size=3, pad_mode='pad', padding=1, stride=2, has_bias=False),
-            nn.BatchNorm2d(32),
-            nn.ReLU6()
-        )
-        layers = []
-        input_channel = 32
-        for t, c, n, s in self.configs:
-            for i in range(n):
-                stride = s if i == 0 else 1
-                layers.append(InvertedBlock(input_channel, c, expand_ratio=t, stride=stride))
-                input_channel = c
-        self.layers = nn.SequentialCell(*layers)
-
-        self.last_conv = nn.SequentialCell(
-            nn.Conv2d(input_channel, 1280, kernel_size=1, padding=0, stride=1, has_bias=False),
-            nn.BatchNorm2d(1280),
-            nn.ReLU6()
-        )
-
-        self.classifier = nn.SequentialCell(
-            nn.Dropout2d(p=0.2),
-            nn.Dense(1280, num_classes)
-        )
-
-        self.avg_pool = nn.AdaptiveAvgPool2d(1)
-
-    def construct(self, x):
-        x = self.stem_conv(x)
-        x = self.layers(x)
-        x = self.last_conv(x)
-        x = self.avg_pool(x).view(-1, 1280)
-        x = self.classifier(x)
-        return x
-
-
-if __name__ == '__main__':
-    dummy_input = ms.ops.randn((1, 3, 224, 224))
-    net = MobileNetV2(in_chans=3, num_classes=1000)
-    output = net(dummy_input)
-    print(output.shape)