ShuffleNetV2 结构（附源码）

本文不细看paper，只看网络结构和源码实现。

看下ShuffleNetV2的结构吧。

image是3通道进去，经过conv1和maxpool,
然后stage2~4则是主题，里面stride = 2和 stride = 1的shuffleBlock分别重复几次。

shuffleBlock如下，左边是stride = 1的，右边是stride = 2的。
可以看到下面左右两个结构的区别，

1. branch1的不同，

2. branch2的第2层：3x3 DWConv,左边的结构中它是stride=1,右边的结构中它是stride=2,
   这就意味着经过左边的block特征图的尺寸会保持不变；经过右边的block, 特征图尺寸会/2,channel会按照设计的增加。

3. stride=1的结构只能通过branch2变化channel, 而且branch1和branch2只会得到一部分channel的x（有channel split）.
   而stride=2的结构中，branch1和branch2都可以得到完整的x，都可以扩充channel。

整体下来，经过stride=1的block后，特征图会保持原尺寸，
而经过stride = 2的block后，特征图的尺寸会/2，channel按照每个stage表格中的变化。

这也就是为什么stride=2只有一层，而stride=1的block可以重复多次。

举个栗子，stage2的in_channel为24, out_channel为116,
每个block是有2个branch的，这个channel要分配一下，比如左边右边各58，经过最后的Concat, 就是116.
stride = 2时channel会增加，特征图尺寸会减半，而stride = 1时channel不变，特征图尺寸不变。

class ShuffleNetV2(nn.Module):
    def __init__(
        self,
        model_size="1.5x",
        out_stages=(2, 3, 4),
        with_last_conv=False,
        kernal_size=3,
        activation="ReLU",
        pretrain=True,
    ):
        super(ShuffleNetV2, self).__init__()
        # out_stages can only be a subset of (2, 3, 4)
        assert set(out_stages).issubset((2, 3, 4))

        print("model size is ", model_size)  #1.0x

        self.stage_repeats = [4, 8, 4]
        self.model_size = model_size
        self.out_stages = out_stages
        self.with_last_conv = with_last_conv #False
        self.kernal_size = kernal_size
        self.activation = activation #LeakyReLU
        if model_size == "0.5x":
            self._stage_out_channels = [24, 48, 96, 192, 1024]
        elif model_size == "1.0x":
            self._stage_out_channels = [24, 116, 232, 464, 1024] 
        elif model_size == "1.5x":
            self._stage_out_channels = [24, 176, 352, 704, 1024]
        elif model_size == "2.0x":
            self._stage_out_channels = [24, 244, 488, 976, 2048]
        else:
            raise NotImplementedError

        # building first layer
        input_channels = 3
        output_channels = self._stage_out_channels[0]  #24
        #conv3x3,s=2.3->24,p=1,BN,LeakyReLU
        self.conv1 = nn.Sequential(
            nn.Conv2d(input_channels, output_channels, 3, 2, 1, bias=False),
            nn.BatchNorm2d(output_channels),
            act_layers(activation),
        )
        input_channels = output_channels  #24

        self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)

        stage_names = ["stage{}".format(i) for i in [2, 3, 4]] #paper中的stage2~4
        #zip是把几个数组的元素打包，以最短的数组为基准
        for name, repeats, output_channels in zip(
            stage_names, self.stage_repeats, self._stage_out_channels[1:]
        ):
        #看paper中的表格，stride=2的repeat一次，stride=1的repeat多少次
            seq = [
                ShuffleV2Block(
                    input_channels, output_channels, 2, activation=activation
                )
            ]
            for i in range(repeats - 1):
                seq.append(
                    ShuffleV2Block(
                        output_channels, output_channels, 1, activation=activation
                    )
                )
            setattr(self, name, nn.Sequential(*seq))
            input_channels = output_channels
        output_channels = self._stage_out_channels[-1]
        if self.with_last_conv:
            conv5 = nn.Sequential(
                nn.Conv2d(input_channels, output_channels, 1, 1, 0, bias=False),
                nn.BatchNorm2d(output_channels),
                act_layers(activation),
            )
            self.stage4.add_module("conv5", conv5)
        self._initialize_weights(pretrain)

    def forward(self, x):
        x = self.conv1(x)
        x = self.maxpool(x)
        output = []
        for i in range(2, 5):
            stage = getattr(self, "stage{}".format(i))
            x = stage(x)
            if i in self.out_stages:
                output.append(x)
        return tuple(output)

ShuffleNetV2 block

class ShuffleV2Block(nn.Module):
    def __init__(self, inp, oup, stride, activation="ReLU"):
        super(ShuffleV2Block, self).__init__()

        if not (1 <= stride <= 3):
            raise ValueError("illegal stride value")
        self.stride = stride

        branch_features = oup // 2  #每个branch分配一半的channel
        assert (self.stride != 1) or (inp == branch_features << 1)

        if self.stride > 1:
            self.branch1 = nn.Sequential(
                self.depthwise_conv(
                    inp, inp, kernel_size=3, stride=self.stride, padding=1
                ),
                nn.BatchNorm2d(inp),
                nn.Conv2d(
                    inp, branch_features, kernel_size=1, stride=1, padding=0, bias=False
                ),
                nn.BatchNorm2d(branch_features),
                act_layers(activation),
            )
        else:
            self.branch1 = nn.Sequential()

        self.branch2 = nn.Sequential(
            nn.Conv2d(
                inp if (self.stride > 1) else branch_features,
                branch_features,
                kernel_size=1,
                stride=1,
                padding=0,
                bias=False,
            ),
            nn.BatchNorm2d(branch_features),
            act_layers(activation),
            self.depthwise_conv(
                branch_features,
                branch_features,
                kernel_size=3,
                stride=self.stride,
                padding=1,
            ),
            nn.BatchNorm2d(branch_features),
            nn.Conv2d(
                branch_features,
                branch_features,
                kernel_size=1,
                stride=1,
                padding=0,
                bias=False,
            ),
            nn.BatchNorm2d(branch_features),
            act_layers(activation),
        )

    @staticmethod
    def depthwise_conv(i, o, kernel_size, stride=1, padding=0, bias=False):
        return nn.Conv2d(i, o, kernel_size, stride, padding, bias=bias, groups=i)

    def forward(self, x):
        if self.stride == 1:
            x1, x2 = x.chunk(2, dim=1) #在dim=1(channel)上分成2块
            out = torch.cat((x1, self.branch2(x2)), dim=1)
        else:
            out = torch.cat((self.branch1(x), self.branch2(x)), dim=1)

        out = channel_shuffle(out, 2)

        return out

channel shuffle在这篇文章里面说过。
channel数转为矩阵，矩阵转置再压平。

def channel_shuffle(x, groups):
    # type: (torch.Tensor, int) -> torch.Tensor
    batchsize, num_channels, height, width = x.data.size()
    channels_per_group = num_channels // groups

    # reshape
    #channel变为groups x channels_per_group
    x = x.view(batchsize, groups, channels_per_group, height, width)

	#转置为channels_per_group x groups
    x = torch.transpose(x, 1, 2).contiguous()

    # flatten
    x = x.view(batchsize, -1, height, width)

    return x