1. 结构图

Input Shape : (3, 7, 7) — Output Shape : (2, 3, 3) — K : (3, 3) — P : (1, 1) — S : (2, 2) — D : (2, 2) — G : 1 

The parts of this post will be divided according to the following arguments. These arguments can be found in the Pytorch documentation of the Conv2d module :

• in_channels (int) — Number of channels in the input image
• out_channels (int) — Number of channels produced by the convolution
• kernel_size (int or tuple) — Size of the convolving kernel
• stride (int or tuple, optional) — Stride of the convolution. Default: 1
• padding (int or tuple, optional) — Zero-padding added to both sides of the input. Default: 0
• dilation (int or tuple, optional) — Spacing between kernel elements. Default: 1
• groups (int, optional) — Number of blocked connections from input channels to output channels. Default: 1
• bias (bool, optional) — If True, adds a learnable bias to the output. Default: True

1.  在Fig 1中, the last layer  is called the translation layer. this layer is responsible for the reducing of the dimension (如图 2, pooling)
2. 尽管该技术可以阻断identity backpropagation, 但是这仅仅发生在DenseNet block中的部分layers, 并不影响梯度流
3. DenseNet 的代码结构主要分为以下三块

                #. DenseLayer: 其主要在 DenseBlock 中完成一个单个的layer

                #. DenseBlock:

               #. TransitionLayer

2. 建立模型

O= Size (width) of output image.
I= Size (width) of input image.
N= Size (width) of kernels used in the Conv Layer.
K= Number of kernels.
S= Stride of the convolution operation.
P= Padding.

The size O of the output image is given by

class _DenseLayer(nn.Module):
    def __init__(
        self, num_input_features: int, growth_rate: int, bn_size: int, drop_rate: float, memory_efficient: bool = False
    ) -> None:
        super().__init__()
        self.norm1 = nn.BatchNorm2d(num_input_features)
        self.relu1 = nn.ReLU(inplace=True)
        self.conv1 = nn.Conv2d(num_input_features, bn_size * growth_rate, kernel_size=1, stride=1, bias=False)

        self.norm2 = nn.BatchNorm2d(bn_size * growth_rate)
        self.relu2 = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(bn_size * growth_rate, growth_rate, kernel_size=3, stride=1, padding=1, bias=False)

        self.drop_rate = float(drop_rate)
        self.memory_efficient = memory_efficient

    def bn_function(self, inputs: List[Tensor]) -> Tensor:
        concated_features = torch.cat(inputs, 1)
        bottleneck_output = self.conv1(self.relu1(self.norm1(concated_features)))  # noqa: T484
        return bottleneck_output

    # todo: rewrite when torchscript supports any
    def any_requires_grad(self, input: List[Tensor]) -> bool:
        for tensor in input:
            if tensor.requires_grad:
                return True
        return False

    @torch.jit.unused  # noqa: T484
    def call_checkpoint_bottleneck(self, input: List[Tensor]) -> Tensor:
        def closure(*inputs):
            return self.bn_function(inputs)

        return cp.checkpoint(closure, *input)

    @torch.jit._overload_method  # noqa: F811
    def forward(self, input: List[Tensor]) -> Tensor:  # noqa: F811
        pass

    @torch.jit._overload_method  # noqa: F811
    def forward(self, input: Tensor) -> Tensor:  # noqa: F811
        pass

    # torchscript does not yet support *args, so we overload method
    # allowing it to take either a List[Tensor] or single Tensor
    def forward(self, input: Tensor) -> Tensor:  # noqa: F811
        if isinstance(input, Tensor):
            prev_features = [input]
        else:
            prev_features = input

        if self.memory_efficient and self.any_requires_grad(prev_features):
            if torch.jit.is_scripting():
                raise Exception("Memory Efficient not supported in JIT")

            bottleneck_output = self.call_checkpoint_bottleneck(prev_features)
        else:
            bottleneck_output = self.bn_function(prev_features)

        new_features = self.conv2(self.relu2(self.norm2(bottleneck_output)))
        if self.drop_rate > 0:
            new_features = F.dropout(new_features, p=self.drop_rate, training=self.training)
        return new_features


class _DenseBlock(nn.ModuleDict):
    _version = 2

    def __init__(
        self,
        num_layers: int,
        num_input_features: int,
        bn_size: int,
        growth_rate: int,
        drop_rate: float,
        memory_efficient: bool = False,
    ) -> None:
        super().__init__()
        for i in range(num_layers):
            layer = _DenseLayer(
                num_input_features + i * growth_rate,
                growth_rate=growth_rate,
                bn_size=bn_size,
                drop_rate=drop_rate,
                memory_efficient=memory_efficient,
            )
            self.add_module("denselayer%d" % (i + 1), layer)

    def forward(self, init_features: Tensor) -> Tensor:
        features = [init_features]
        for name, layer in self.items():
            new_features = layer(features)
            features.append(new_features)
        return torch.cat(features, 1)


class _Transition(nn.Sequential):
    def __init__(self, num_input_features: int, num_output_features: int) -> None:
        super().__init__()
        self.norm = nn.BatchNorm2d(num_input_features)
        self.relu = nn.ReLU(inplace=True)
        self.conv = nn.Conv2d(num_input_features, num_output_features, kernel_size=1, stride=1, bias=False)
        self.pool = nn.AvgPool2d(kernel_size=2, stride=2)


class DenseNet(nn.Module):
    r"""Densenet-BC model class, based on
    `"Densely Connected Convolutional Networks" <https://arxiv.org/pdf/1608.06993.pdf>`_.

    Args:
        growth_rate (int) - how many filters to add each layer (`k` in paper)
        block_config (list of 4 ints) - how many layers in each pooling block
        num_init_features (int) - the number of filters to learn in the first convolution layer
        bn_size (int) - multiplicative factor for number of bottle neck layers
          (i.e. bn_size * k features in the bottleneck layer)
        drop_rate (float) - dropout rate after each dense layer
        num_classes (int) - number of classification classes
        memory_efficient (bool) - If True, uses checkpointing. Much more memory efficient,
          but slower. Default: *False*. See `"paper" <https://arxiv.org/pdf/1707.06990.pdf>`_.
    """

    def __init__(
        self,
        growth_rate: int = 32,
        block_config: Tuple[int, int, int, int] = (6, 12, 24, 16),
        num_init_features: int = 64,
        bn_size: int = 4,
        drop_rate: float = 0,
        num_classes: int = 1000,
        memory_efficient: bool = False,
    ) -> None:

        super().__init__()
        # _log_api_usage_once(self)

        # First convolution
        self.features = nn.Sequential(
            OrderedDict(
                [
                    ("conv0", nn.Conv2d(3, num_init_features, kernel_size=7, stride=2, padding=3, bias=False)),
                    ("norm0", nn.BatchNorm2d(num_init_features)),
                    ("relu0", nn.ReLU(inplace=True)),
                    ("pool0", nn.MaxPool2d(kernel_size=3, stride=2, padding=1)),
                ]
            )
        )

        # Each denseblock
        num_features = num_init_features
        for i, num_layers in enumerate(block_config):
            block = _DenseBlock(
                num_layers=num_layers,
                num_input_features=num_features,
                bn_size=bn_size,
                growth_rate=growth_rate,
                drop_rate=drop_rate,
                memory_efficient=memory_efficient,
            )
            self.features.add_module("denseblock%d" % (i + 1), block)

            num_features = num_features + num_layers * growth_rate

            if i != len(block_config) - 1:
                trans = _Transition(num_input_features=num_features, num_output_features=num_features // 2)
                # self.features.add_module("transition%d" % (i + 1), trans)
                # num_features = num_features // 2

        # Final batch norm
        self.features.add_module("norm5", nn.BatchNorm2d(num_features))

        # Linear layer
        self.classifier = nn.Linear(num_features, num_classes)

        # Official init from torch repo.
        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight)
            elif isinstance(m, nn.BatchNorm2d):
                nn.init.constant_(m.weight, 1)
                nn.init.constant_(m.bias, 0)
            elif isinstance(m, nn.Linear):
                nn.init.constant_(m.bias, 0)

    def forward(self, x: Tensor) -> Tensor:
        features = self.features(x)
        out = F.relu(features, inplace=True)
        out = F.adaptive_avg_pool2d(out, (1, 1))
        out = torch.flatten(out, 1)
        out = self.classifier(out)
        return out