mobilenet v2笔记

mobilenet v2

相对于mobilenet v1来说,其v2改进的地方在于:

• 像resnet一样加入了residual connection高速通道,增加对图像高层语义信息与低纬特征融合
• Linear Bottlenecks,通过不同通道数对relu6激活函数分析

Linear Bottlenecks

本身意义上说:relu函数只取输入的非负值，且认为这个行为导致了信息损失

流形学习

文章假设了每一层的特征图激活值形成了一种未知流形，虽然无法直接知道这个流形是什么，但是流形学习中最基本的想法就是高维数据能够映射到低维空间，因而这些激活值构成的流形也是能映射的。在神经网络中这种映射就体现在降低特征图的channel数，类似MobileNet v1中设置了一个超参数宽度乘子α，但是问题是这个宽度因子如何选择？降低到多少channel数能够最大程度的保留所有的信息。映射到一个合适的低维空间即等价于通过这个乘子降低到合适的channel数，且作者认为在宽度乘子作用下能够使得激活值形成的低维流形能够占满整个低维空间。就是在这个条件之下，文章通过公式证明了只有在激活值能够得到输入空间的一个低维流形时，relu函数才能够保留完备的信息。

这个性质表明了激活值的channel数量较少时，relu无法保留完备的信息，会丢失很多信息，所以这种情况下不适合用relu，如下图所示，将原始二维信息，经过一个矩阵映射到高维空间中，再经过一个relu，反向映射回原始二维空间，由于原始的二维空间是n=15,30的低维流形，而不是n=2,3的低维流形，所以在n=2,3时信息丢失较多。

image

基于以上结论,作者设计了linear bottleneck layer,其有三个卷积操作:

• (1) 1X1 的PointwiseConvolution,用以提升通道数(后接激活函数relu6)
• (2) 3X3 的DepthwiseConvolution,正常卷积抽特征处理(后接激活函数relu6)
• (3)1X1 的PointwiseConvolution将channel数量降下来(不接激活函数relu6)

Bottleneck就体现在最后的这个降低channel数的PointwiseConvolution，根据之前对relu的研究，这种维数较低的channel之后再加relu会损失信息，所以linear则体现在最后的PointwiseConvolution之后不像之前一样接relu了。

image

linear bottleneck本质上是不带ReLU的1x1的卷积层。

ResNet block

image

传统的resnet block :先用1x1降通道过ReLU，再3x3空间卷积过ReLU，再用1x1卷积过ReLU恢复通道，并和输入相加。之所以要1x1卷积降通道，是为了减少计算量，不然中间的3x3空间卷积计算量太大。沙漏型,两边宽中间窄.

现在我们中间的3x3卷积变为了Depthwise的了，计算量很少了，所以通道可以多一点，效果更好，所以通过1x1卷积先提升通道数，再用Depthwise的3x3空间卷积，再用1x1卷积降低维度。两端的通道数都很小，所以1x1卷积升通道或降通道计算量都并不大，而中间通道数虽然多，但是是Depthwise的卷积，计算量也不大。作者称之为Inverted residual block，两边窄中间宽，像柳叶，较小的计算量得到较好的性能。

所以mobilenet v2的block 如图所示:

image

代码:

from torch import nn
from .utils import load_state_dict_from_url


__all__ = ['MobileNetV2', 'mobilenet_v2']


model_urls = {
    'mobilenet_v2': 'https://download.pytorch.org/models/mobilenet_v2-b0353104.pth',
}


def _make_divisible(v, divisor, min_value=None):
    """
    This function is taken from the original tf repo.
    It ensures that all layers have a channel number that is divisible by 8
    It can be seen here:
    https://github.com/tensorflow/models/blob/master/research/slim/nets/mobilenet/mobilenet.py
    :param v:
    :param divisor:
    :param min_value:
    :return:
    """
    if min_value is None:
        min_value = divisor
    new_v = max(min_value, int(v + divisor / 2) // divisor * divisor)
    # Make sure that round down does not go down by more than 10%.
    if new_v < 0.9 * v:
        new_v += divisor
    return new_v


class ConvBNReLU(nn.Sequential):
    def __init__(self, in_planes, out_planes, kernel_size=3, stride=1, groups=1):
        padding = (kernel_size - 1) // 2
        super(ConvBNReLU, self).__init__(
            nn.Conv2d(in_planes, out_planes, kernel_size, stride, padding, groups=groups, bias=False),
            nn.BatchNorm2d(out_planes),
            nn.ReLU6(inplace=True)
        )


class InvertedResidual(nn.Module):
    def __init__(self, inp, oup, stride, expand_ratio):
        super(InvertedResidual, self).__init__()
        self.stride = stride
        assert stride in [1, 2]

        hidden_dim = int(round(inp * expand_ratio))
        self.use_res_connect = self.stride == 1 and inp == oup

        layers = []
        if expand_ratio != 1:
            # pw
            layers.append(ConvBNReLU(inp, hidden_dim, kernel_size=1))
        layers.extend([
            # dw
            ConvBNReLU(hidden_dim, hidden_dim, stride=stride, groups=hidden_dim),
            # pw-linear
            nn.Conv2d(hidden_dim, oup, 1, 1, 0, bias=False),
            nn.BatchNorm2d(oup),
        ])
        self.conv = nn.Sequential(*layers)

    def forward(self, x):
        if self.use_res_connect:
            return x + self.conv(x)
        else:
            return self.conv(x)


class MobileNetV2(nn.Module):
    def __init__(self,
                 num_classes=1000,
                 width_mult=1.0,
                 inverted_residual_setting=None,
                 round_nearest=8,
                 block=None):
        """
        MobileNet V2 main class
        Args:
            num_classes (int): Number of classes
            width_mult (float): Width multiplier - adjusts number of channels in each layer by this amount
            inverted_residual_setting: Network structure
            round_nearest (int): Round the number of channels in each layer to be a multiple of this number
            Set to 1 to turn off rounding
            block: Module specifying inverted residual building block for mobilenet
        """
        super(MobileNetV2, self).__init__()

        if block is None:
            block = InvertedResidual
        input_channel = 32
        last_channel = 1280

        if inverted_residual_setting is None:
            inverted_residual_setting = [
                # 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],
            ]

        # only check the first element, assuming user knows t,c,n,s are required
        if len(inverted_residual_setting) == 0 or len(inverted_residual_setting[0]) != 4:
            raise ValueError("inverted_residual_setting should be non-empty "
                             "or a 4-element list, got {}".format(inverted_residual_setting))

        # building first layer
        input_channel = _make_divisible(input_channel * width_mult, round_nearest)
        self.last_channel = _make_divisible(last_channel * max(1.0, width_mult), round_nearest)
        features = [ConvBNReLU(3, input_channel, stride=2)]
        # building inverted residual blocks
        for t, c, n, s in inverted_residual_setting:
            output_channel = _make_divisible(c * width_mult, round_nearest)
            for i in range(n):
                stride = s if i == 0 else 1
                features.append(block(input_channel, output_channel, stride, expand_ratio=t))
                input_channel = output_channel
        # building last several layers
        features.append(ConvBNReLU(input_channel, self.last_channel, kernel_size=1))
        # make it nn.Sequential
        self.features = nn.Sequential(*features)

        # building classifier
        self.classifier = nn.Sequential(
            nn.Dropout(0.2),
            nn.Linear(self.last_channel, num_classes),
        )

        # weight initialization
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode='fan_out')
                if m.bias is not None:
                    nn.init.zeros_(m.bias)
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.ones_(m.weight)
                nn.init.zeros_(m.bias)
            elif isinstance(m, nn.Linear):
                nn.init.normal_(m.weight, 0, 0.01)
                nn.init.zeros_(m.bias)

    def _forward_impl(self, x):
        # This exists since TorchScript doesn't support inheritance, so the superclass method
        # (this one) needs to have a name other than `forward` that can be accessed in a subclass
        x = self.features(x)
        x = x.mean([2, 3])
        x = self.classifier(x)
        return x

    def forward(self, x):
        return self._forward_impl(x)


def mobilenet_v2(pretrained=False, progress=True, **kwargs):
    """
    Constructs a MobileNetV2 architecture from
    `"MobileNetV2: Inverted Residuals and Linear Bottlenecks" <https://arxiv.org/abs/1801.04381>`_.
    Args:
        pretrained (bool): If True, returns a model pre-trained on ImageNet
        progress (bool): If True, displays a progress bar of the download to stderr
    """
    model = MobileNetV2(**kwargs)
    if pretrained:
        state_dict = load_state_dict_from_url(model_urls['mobilenet_v2'],
                                              progress=progress)
        model.load_state_dict(state_dict)
    return model