博客

本文转载自知乎

https://zhuanlan.zhihu.com/p/68531147?utm_source=qq

，侵删

本文是对 DeepLab 系列的概括，主要讨论模型的设计和改进，附 Pytorch 实现代码，略去训练细节以及性能细节，这些都可以在原论文中找到。

原论文地址：

DeepLabv1


https://arxiv.org/pdf/1412.7062v3.pdf


DeepLabv2


DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs


DeepLabv3


Rethinking Atrous Convolution for Semantic Image Segmentation


DeepLabv3+


Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation

DeepLabv1

DeepLabv1 模型结构很容易理解：

1. 首先它是个 VGG-16
2. 然后为了使图像语义分割更准确，5 个 max-pooling 层 skip 了后两个（具体实现上，看G站上的代码，似乎没有去除，而是保留了后两个 max-pooling ，只是将 stride = 2 改为 stride = 1，kernal = 3），最后卷积层的输出整体 stride 从 32x 下降至 8x。
3. 参考
   
   Uno Whoiam：空洞卷积（Dilated Convolution）：有之以为利，无之以为用
   
   ，由于后两个 max-pooling 影响了其后的卷积层，使其视野分别下降了 2x 和 4x，为了保持其原来的视野，便将其改成空洞卷积，dilation 分别为 2 和 4，理念与DRN一致：

4. 当然，它也是一个全卷积网络

Uno Whoiam：FCN：从图片分类到像素分类

，即将全连接层替换成

的卷积层，输出和原图大小一致的特征图，对每个像素分类。

5. 使用双线性插值上采样 8x 得到和原图大小一致的像素分类图。

6. 使用 CRF（条件随机场）使最后分类结果的边缘更加精细：

啥是 CRF 呢？这里只给论文里的公式，不深究，v3 以及之后就没用这玩意了：

其中

为 DCNN 输出的置信度；

，p 表示像素的位置，I 表示像素的 RGB 数值；如何理解这玩意呢？简单来说就是在对一个像素做分类时，不光考虑 DCNN 输出的结果，还要考虑周围像素的意见尤其像素值比较接近的，这样得出的语义分割结果会有更好的边缘。

7. 多尺寸预测，希望获得更好的边界信息，与FCN skip layer类似，具体实现上，在输入图片与前四个 max pooling 后添加

和

的卷积层，这四个预测结果与最终模型输出拼接（concatenate）到一起，相当于多了128*5=640个channel。虽然效果不如dense CRF，但也有一定提高。最终模型是结合了Desne CRF与Multi-scale Prediction。

一个简单的 Pytorch 实现如下，使用 ResNet，第一层为

普通卷积，stride = 2，紧跟着 stride = 2 的 max-pooling，尔后一个普通的 bottleneck ，一个 stride = 2 的 bottleneck，然后 dilation =2、dilation =4 的bottleneck。

参考链接：

kazuto1011/deeplab-pytorch

from __future__ import absolute_import, print_function

import torch
import torch.nn as nn
import torch.nn.functional as F


class DeepLabV1(nn.Sequential):
    """
    DeepLab v1: Dilated ResNet + 1x1 Conv
    Note that this is just a container for loading the pretrained COCO model and not mentioned as "v1" in papers.
    """

    def __init__(self, n_classes, n_blocks):
        super(DeepLabV1, self).__init__()
        ch = [64 * 2 ** p for p in range(6)]
        self.add_module("layer1", _Stem(ch[0]))
        self.add_module("layer2", _ResLayer(n_blocks[0], ch[0], ch[2], 1, 1))
        self.add_module("layer3", _ResLayer(n_blocks[1], ch[2], ch[3], 2, 1))
        self.add_module("layer4", _ResLayer(n_blocks[2], ch[3], ch[4], 1, 2))
        self.add_module("layer5", _ResLayer(n_blocks[3], ch[4], ch[5], 1, 4))
        self.add_module("fc", nn.Conv2d(2048, n_classes, 1))


try:
    from encoding.nn import SyncBatchNorm

    _BATCH_NORM = SyncBatchNorm
except:
    _BATCH_NORM = nn.BatchNorm2d

_BOTTLENECK_EXPANSION = 4


class _ConvBnReLU(nn.Sequential):
    """
    Cascade of 2D convolution, batch norm, and ReLU.
    """

    BATCH_NORM = _BATCH_NORM

    def __init__(
            self, in_ch, out_ch, kernel_size, stride, padding, dilation, relu=True
    ):
        super(_ConvBnReLU, self).__init__()
        self.add_module(
            "conv",
            nn.Conv2d(
                in_ch, out_ch, kernel_size, stride, padding, dilation, bias=False
            ),
        )
        self.add_module("bn", _BATCH_NORM(out_ch, eps=1e-5, momentum=0.999))

        if relu:
            self.add_module("relu", nn.ReLU())


class _Bottleneck(nn.Module):
    """
    Bottleneck block of MSRA ResNet.
    """

    def __init__(self, in_ch, out_ch, stride, dilation, downsample):
        super(_Bottleneck, self).__init__()
        mid_ch = out_ch // _BOTTLENECK_EXPANSION
        self.reduce = _ConvBnReLU(in_ch, mid_ch, 1, stride, 0, 1, True)
        self.conv3x3 = _ConvBnReLU(mid_ch, mid_ch, 3, 1, dilation, dilation, True)
        self.increase = _ConvBnReLU(mid_ch, out_ch, 1, 1, 0, 1, False)
        self.shortcut = (
            _ConvBnReLU(in_ch, out_ch, 1, stride, 0, 1, False)
            if downsample
            else lambda x: x  # identity
        )

    def forward(self, x):
        h = self.reduce(x)
        h = self.conv3x3(h)
        h = self.increase(h)
        h += self.shortcut(x)
        return F.relu(h)


class _ResLayer(nn.Sequential):
    """
    Residual layer with multi grids
    """

    def __init__(self, n_layers, in_ch, out_ch, stride, dilation, multi_grids=None):
        super(_ResLayer, self).__init__()

        if multi_grids is None:
            multi_grids = [1 for _ in range(n_layers)]
        else:
            assert n_layers == len(multi_grids)

        # Downsampling is only in the first block
        for i in range(n_layers):
            self.add_module(
                "block{}".format(i + 1),
                _Bottleneck(
                    in_ch=(in_ch if i == 0 else out_ch),
                    out_ch=out_ch,
                    stride=(stride if i == 0 else 1),
                    dilation=dilation * multi_grids[i],
                    downsample=(True if i == 0 else False),
                ),
            )


class _Stem(nn.Sequential):
    """
    The 1st conv layer.
    Note that the max pooling is different from both MSRA and FAIR ResNet.
    """

    def __init__(self, out_ch):
        super(_Stem, self).__init__()
        self.add_module("conv1", _ConvBnReLU(3, out_ch, 7, 2, 3, 1))
        self.add_module("pool", nn.MaxPool2d(3, 2, 1, ceil_mode=True))


class _Flatten(nn.Module):
    def forward(self, x):
        return x.view(x.size(0), -1)


if __name__ == "__main__":
    model = DeepLabV1(n_classes=21, n_blocks=[3, 4, 23, 3])
    model.eval()
    image = torch.randn(1, 3, 513, 513)

    print(model)
    print("input:", image.shape)
    print("output:", model(image).shape)

DeepLabv2

DeepLabv2 相对于 v1 最大的改动是增加了受

SPP（Spacial Pyramid Pooling）

启发得来的 ASPP（Atrous Spacial Pyramid Pooling），在模型最后进行像素分类之前增加一个类似 Inception 的结构，包含不同 rate（空洞间隔） 的 Atrous Conv（空洞卷积），增强模型识别不同尺寸的同一物体的能力：

DeepLabv2 Pytorch 实现：

from __future__ import absolute_import, print_function

import torch
import torch.nn as nn
import torch.nn.functional as F


class _ASPP(nn.Module):
    """
    Atrous spatial pyramid pooling (ASPP)
    """

    def __init__(self, in_ch, out_ch, rates):
        super(_ASPP, self).__init__()
        for i, rate in enumerate(rates):
            self.add_module(
                "c{}".format(i),
                nn.Conv2d(in_ch, out_ch, 3, 1, padding=rate, dilation=rate, bias=True),
            )

        for m in self.children():
            nn.init.normal_(m.weight, mean=0, std=0.01)
            nn.init.constant_(m.bias, 0)

    def forward(self, x):
        return sum([stage(x) for stage in self.children()])


class DeepLabV2(nn.Sequential):
    """
    DeepLab v2: Dilated ResNet + ASPP
    Output stride is fixed at 8
    """

    def __init__(self, n_classes, n_blocks, atrous_rates):
        super(DeepLabV2, self).__init__()
        ch = [64 * 2 ** p for p in range(6)]
        self.add_module("layer1", _Stem(ch[0]))
        self.add_module("layer2", _ResLayer(n_blocks[0], ch[0], ch[2], 1, 1))
        self.add_module("layer3", _ResLayer(n_blocks[1], ch[2], ch[3], 2, 1))
        self.add_module("layer4", _ResLayer(n_blocks[2], ch[3], ch[4], 1, 2))
        self.add_module("layer5", _ResLayer(n_blocks[3], ch[4], ch[5], 1, 4))
        self.add_module("aspp", _ASPP(ch[5], n_classes, atrous_rates))

    def freeze_bn(self):
        for m in self.modules():
            if isinstance(m, _ConvBnReLU.BATCH_NORM):
                m.eval()


try:
    from encoding.nn import SyncBatchNorm

    _BATCH_NORM = SyncBatchNorm
except:
    _BATCH_NORM = nn.BatchNorm2d

_BOTTLENECK_EXPANSION = 4



class _ConvBnReLU(nn.Sequential):
    """
    Cascade of 2D convolution, batch norm, and ReLU.
    """

    BATCH_NORM = _BATCH_NORM

    def __init__(
            self, in_ch, out_ch, kernel_size, stride, padding, dilation, relu=True
    ):
        super(_ConvBnReLU, self).__init__()
        self.add_module(
            "conv",
            nn.Conv2d(
                in_ch, out_ch, kernel_size, stride, padding, dilation, bias=False
            ),
        )
        self.add_module("bn", _BATCH_NORM(out_ch, eps=1e-5, momentum=0.999))

        if relu:
            self.add_module("relu", nn.ReLU())


class _Bottleneck(nn.Module):
    """
    Bottleneck block of MSRA ResNet.
    """

    def __init__(self, in_ch, out_ch, stride, dilation, downsample):
        super(_Bottleneck, self).__init__()
        mid_ch = out_ch // _BOTTLENECK_EXPANSION
        self.reduce = _ConvBnReLU(in_ch, mid_ch, 1, stride, 0, 1, True)
        self.conv3x3 = _ConvBnReLU(mid_ch, mid_ch, 3, 1, dilation, dilation, True)
        self.increase = _ConvBnReLU(mid_ch, out_ch, 1, 1, 0, 1, False)
        self.shortcut = (
            _ConvBnReLU(in_ch, out_ch, 1, stride, 0, 1, False)
            if downsample
            else lambda x: x  # identity
        )

    def forward(self, x):
        h = self.reduce(x)
        h = self.conv3x3(h)
        h = self.increase(h)
        h += self.shortcut(x)
        return F.relu(h)


class _ResLayer(nn.Sequential):
    """
    Residual layer with multi grids
    """

    def __init__(self, n_layers, in_ch, out_ch, stride, dilation, multi_grids=None):
        super(_ResLayer, self).__init__()

        if multi_grids is None:
            multi_grids = [1 for _ in range(n_layers)]
        else:
            assert n_layers == len(multi_grids)

        # Downsampling is only in the first block
        for i in range(n_layers):
            self.add_module(
                "block{}".format(i + 1),
                _Bottleneck(
                    in_ch=(in_ch if i == 0 else out_ch),
                    out_ch=out_ch,
                    stride=(stride if i == 0 else 1),
                    dilation=dilation * multi_grids[i],
                    downsample=(True if i == 0 else False),
                ),
            )


class _Stem(nn.Sequential):
    """
    The 1st conv layer.
    Note that the max pooling is different from both MSRA and FAIR ResNet.
    """

    def __init__(self, out_ch):
        super(_Stem, self).__init__()
        self.add_module("conv1", _ConvBnReLU(3, out_ch, 7, 2, 3, 1))
        self.add_module("pool", nn.MaxPool2d(3, 2, 1, ceil_mode=True))


if __name__ == "__main__":
    model = DeepLabV2(
        n_classes=21, n_blocks=[3, 4, 23, 3], atrous_rates=[6, 12, 18, 24]
    )
    model.eval()
    image = torch.randn(1, 3, 513, 513)

    print(model)
    print("input:", image.shape)
    print("output:", model(image).shape)

另外，DeepLabv2 采用了 Poly 的训练策略：

当

时，模型效果要比普通的分段学习率策略要高

，Pytorch 实现如下：

https://github.com/kazuto1011/deeplab-pytorch/blob/master/libs/utils/lr_scheduler.py

from torch.optim.lr_scheduler import _LRScheduler


class PolynomialLR(_LRScheduler):
    def __init__(self, optimizer, step_size, iter_max, power, last_epoch=-1):
        self.step_size = step_size
        self.iter_max = iter_max
        self.power = power
        super(PolynomialLR, self).__init__(optimizer, last_epoch)

    def polynomial_decay(self, lr):
        return lr * (1 - float(self.last_epoch) / self.iter_max) ** self.power

    def get_lr(self):
        if (
            (self.last_epoch == 0)
            or (self.last_epoch % self.step_size != 0)
            or (self.last_epoch > self.iter_max)
        ):
            return [group["lr"] for group in self.optimizer.param_groups]
        return [self.polynomial_decay(lr) for lr in self.base_lrs]

DeepLabv3

DeepLabv3 的主要变化如下：

1. 使用了Multi-Grid 策略，即在模型后端多加几层不同 rate 的空洞卷积：

2. 将 batch normalization 加入到 ASPP模块.

3. 具有不同 atrous rates 的 ASPP 能够有效的捕获多尺度信息。不过，论文发现，随着sampling rate的增加，有效filter特征权重(即有效特征区域，而不是补零区域的权重)的数量会变小，极端情况下，当空洞卷积的 rate 和 feature map 的大小一致时，

卷积会退化成

：

为了保留较大视野的空洞卷积的同时解决这个问题，DeepLabv3 的 ASPP 加入了 全局池化层+conv1x1+双线性插值上采样 的模块：

DeepLabv3 的Pytorch 实现如下：

from __future__ import absolute_import, print_function

from collections import OrderedDict

import torch
import torch.nn as nn
import torch.nn.functional as F


class _ImagePool(nn.Module):
    def __init__(self, in_ch, out_ch):
        super().__init__()
        self.pool = nn.AdaptiveAvgPool2d(1)
        self.conv = _ConvBnReLU(in_ch, out_ch, 1, 1, 0, 1)

    def forward(self, x):
        _, _, H, W = x.shape
        h = self.pool(x)
        h = self.conv(h)
        h = F.interpolate(h, size=(H, W), mode="bilinear", align_corners=False)
        return h


class _ASPP(nn.Module):
    """
    Atrous spatial pyramid pooling with image-level feature
    """

    def __init__(self, in_ch, out_ch, rates):
        super(_ASPP, self).__init__()
        self.stages = nn.Module()
        self.stages.add_module("c0", _ConvBnReLU(in_ch, out_ch, 1, 1, 0, 1))
        for i, rate in enumerate(rates):
            self.stages.add_module(
                "c{}".format(i + 1),
                _ConvBnReLU(in_ch, out_ch, 3, 1, padding=rate, dilation=rate),
            )
        self.stages.add_module("imagepool", _ImagePool(in_ch, out_ch))

    def forward(self, x):
        return torch.cat([stage(x) for stage in self.stages.children()], dim=1)


class DeepLabV3(nn.Sequential):
    """
    DeepLab v3: Dilated ResNet with multi-grid + improved ASPP
    """

    def __init__(self, n_classes, n_blocks, atrous_rates, multi_grids, output_stride):
        super(DeepLabV3, self).__init__()

        # Stride and dilation
        if output_stride == 8:
            s = [1, 2, 1, 1]
            d = [1, 1, 2, 4]
        elif output_stride == 16:
            s = [1, 2, 2, 1]
            d = [1, 1, 1, 2]

        ch = [64 * 2 ** p for p in range(6)]
        self.add_module("layer1", _Stem(ch[0]))
        self.add_module("layer2", _ResLayer(n_blocks[0], ch[0], ch[2], s[0], d[0]))
        self.add_module("layer3", _ResLayer(n_blocks[1], ch[2], ch[3], s[1], d[1]))
        self.add_module("layer4", _ResLayer(n_blocks[2], ch[3], ch[4], s[2], d[2]))
        self.add_module(
            "layer5", _ResLayer(n_blocks[3], ch[4], ch[5], s[3], d[3], multi_grids)
        )
        self.add_module("aspp", _ASPP(ch[5], 256, atrous_rates))
        concat_ch = 256 * (len(atrous_rates) + 2)
        self.add_module("fc1", _ConvBnReLU(concat_ch, 256, 1, 1, 0, 1))
        self.add_module("fc2", nn.Conv2d(256, n_classes, kernel_size=1))


try:
    from encoding.nn import SyncBatchNorm

    _BATCH_NORM = SyncBatchNorm
except:
    _BATCH_NORM = nn.BatchNorm2d

_BOTTLENECK_EXPANSION = 4


class _ConvBnReLU(nn.Sequential):
    """
    Cascade of 2D convolution, batch norm, and ReLU.
    """

    BATCH_NORM = _BATCH_NORM

    def __init__(
            self, in_ch, out_ch, kernel_size, stride, padding, dilation, relu=True
    ):
        super(_ConvBnReLU, self).__init__()
        self.add_module(
            "conv",
            nn.Conv2d(
                in_ch, out_ch, kernel_size, stride, padding, dilation, bias=False
            ),
        )
        self.add_module("bn", _BATCH_NORM(out_ch, eps=1e-5, momentum=0.999))

        if relu:
            self.add_module("relu", nn.ReLU())


class _Bottleneck(nn.Module):
    """
    Bottleneck block of MSRA ResNet.
    """

    def __init__(self, in_ch, out_ch, stride, dilation, downsample):
        super(_Bottleneck, self).__init__()
        mid_ch = out_ch // _BOTTLENECK_EXPANSION
        self.reduce = _ConvBnReLU(in_ch, mid_ch, 1, stride, 0, 1, True)
        self.conv3x3 = _ConvBnReLU(mid_ch, mid_ch, 3, 1, dilation, dilation, True)
        self.increase = _ConvBnReLU(mid_ch, out_ch, 1, 1, 0, 1, False)
        self.shortcut = (
            _ConvBnReLU(in_ch, out_ch, 1, stride, 0, 1, False)
            if downsample
            else lambda x: x  # identity
        )

    def forward(self, x):
        h = self.reduce(x)
        h = self.conv3x3(h)
        h = self.increase(h)
        h += self.shortcut(x)
        return F.relu(h)


class _ResLayer(nn.Sequential):
    """
    Residual layer with multi grids
    """

    def __init__(self, n_layers, in_ch, out_ch, stride, dilation, multi_grids=None):
        super(_ResLayer, self).__init__()

        if multi_grids is None:
            multi_grids = [1 for _ in range(n_layers)]
        else:
            assert n_layers == len(multi_grids)

        # Downsampling is only in the first block
        for i in range(n_layers):
            self.add_module(
                "block{}".format(i + 1),
                _Bottleneck(
                    in_ch=(in_ch if i == 0 else out_ch),
                    out_ch=out_ch,
                    stride=(stride if i == 0 else 1),
                    dilation=dilation * multi_grids[i],
                    downsample=(True if i == 0 else False),
                ),
            )


class _Stem(nn.Sequential):
    """
    The 1st conv layer.
    Note that the max pooling is different from both MSRA and FAIR ResNet.
    """

    def __init__(self, out_ch):
        super(_Stem, self).__init__()
        self.add_module("conv1", _ConvBnReLU(3, out_ch, 7, 2, 3, 1))
        self.add_module("pool", nn.MaxPool2d(3, 2, 1, ceil_mode=True))


if __name__ == "__main__":
    model = DeepLabV3(
        n_classes=21,
        n_blocks=[3, 4, 23, 3],
        atrous_rates=[6, 12, 18],
        multi_grids=[1, 2, 4],
        output_stride=8,
    )
    model.eval()
    image = torch.randn(1, 3, 513, 513)

    print(model)
    print("input:", image.shape)
    print("output:", model(image).shape)

DeepLabv3+

V3+ 最大的改进是将 DeepLab 的 DCNN 部分看做 Encoder，将 DCNN 输出的特征图上采样成原图大小的部分看做 Decoder ，构成 Encoder+Decoder 体系，双线性插值上采样便是一个简单的 Decoder，而强化 Decoder 便可使模型整体在图像语义分割边缘部分取得良好的结果。

具体来说，DeepLabV3+ 在 stride = 16 的DeepLabv3 模型输出上采样 4x 后，将 DCNN 中 0.25x 的输出使用

的卷积降维后与之连接（concat）再使用

卷积处理后双线性插值上采样 4 倍后得到相对于 DeepLabv3 更精细的结果。

DeepLabv3+的其他改进还有：

1. 借鉴MobileNet，使用 Depth-wise 空洞卷积+
   
   卷积：

2. 使用修改过的 Xception：

使用 Pytorch 的DeepLabv3+ 实现如下：

from __future__ import absolute_import, print_function

from collections import OrderedDict

import torch
import torch.nn as nn
import torch.nn.functional as F


class _ASPP(nn.Module):
    """
    Atrous spatial pyramid pooling with image-level feature
    """

    def __init__(self, in_ch, out_ch, rates):
        super(_ASPP, self).__init__()
        self.stages = nn.Module()
        self.stages.add_module("c0", _ConvBnReLU(in_ch, out_ch, 1, 1, 0, 1))
        for i, rate in enumerate(rates):
            self.stages.add_module(
                "c{}".format(i + 1),
                _ConvBnReLU(in_ch, out_ch, 3, 1, padding=rate, dilation=rate),
            )
        self.stages.add_module("imagepool", _ImagePool(in_ch, out_ch))

    def forward(self, x):
        return torch.cat([stage(x) for stage in self.stages.children()], dim=1)


class DeepLabV3Plus(nn.Module):
    """
    DeepLab v3+: Dilated ResNet with multi-grid + improved ASPP + decoder
    """

    def __init__(self, n_classes, n_blocks, atrous_rates, multi_grids, output_stride):
        super(DeepLabV3Plus, self).__init__()

        # Stride and dilation
        if output_stride == 8:
            s = [1, 2, 1, 1]
            d = [1, 1, 2, 4]
        elif output_stride == 16:
            s = [1, 2, 2, 1]
            d = [1, 1, 1, 2]

        # Encoder
        ch = [64 * 2 ** p for p in range(6)]
        self.layer1 = _Stem(ch[0])
        self.layer2 = _ResLayer(n_blocks[0], ch[0], ch[2], s[0], d[0])
        self.layer3 = _ResLayer(n_blocks[1], ch[2], ch[3], s[1], d[1])
        self.layer4 = _ResLayer(n_blocks[2], ch[3], ch[4], s[2], d[2])
        self.layer5 = _ResLayer(n_blocks[3], ch[4], ch[5], s[3], d[3], multi_grids)
        self.aspp = _ASPP(ch[5], 256, atrous_rates)
        concat_ch = 256 * (len(atrous_rates) + 2)
        self.add_module("fc1", _ConvBnReLU(concat_ch, 256, 1, 1, 0, 1))

        # Decoder
        self.reduce = _ConvBnReLU(256, 48, 1, 1, 0, 1)
        self.fc2 = nn.Sequential(
            OrderedDict(
                [
                    ("conv1", _ConvBnReLU(304, 256, 3, 1, 1, 1)),
                    ("conv2", _ConvBnReLU(256, 256, 3, 1, 1, 1)),
                    ("conv3", nn.Conv2d(256, n_classes, kernel_size=1)),
                ]
            )
        )

    def forward(self, x):
        h = self.layer1(x)
        h = self.layer2(h)
        h_ = self.reduce(h)
        h = self.layer3(h)
        h = self.layer4(h)
        h = self.layer5(h)
        h = self.aspp(h)
        h = self.fc1(h)
        h = F.interpolate(h, size=h_.shape[2:], mode="bilinear", align_corners=False)
        h = torch.cat((h, h_), dim=1)
        h = self.fc2(h)
        h = F.interpolate(h, size=x.shape[2:], mode="bilinear", align_corners=False)
        return h


try:
    from encoding.nn import SyncBatchNorm

    _BATCH_NORM = SyncBatchNorm
except:
    _BATCH_NORM = nn.BatchNorm2d

_BOTTLENECK_EXPANSION = 4


class _ConvBnReLU(nn.Sequential):
    """
    Cascade of 2D convolution, batch norm, and ReLU.
    """

    BATCH_NORM = _BATCH_NORM

    def __init__(
            self, in_ch, out_ch, kernel_size, stride, padding, dilation, relu=True
    ):
        super(_ConvBnReLU, self).__init__()
        self.add_module(
            "conv",
            nn.Conv2d(
                in_ch, out_ch, kernel_size, stride, padding, dilation, bias=False
            ),
        )
        self.add_module("bn", _BATCH_NORM(out_ch, eps=1e-5, momentum=0.999))

        if relu:
            self.add_module("relu", nn.ReLU())


class _Bottleneck(nn.Module):
    """
    Bottleneck block of MSRA ResNet.
    """

    def __init__(self, in_ch, out_ch, stride, dilation, downsample):
        super(_Bottleneck, self).__init__()
        mid_ch = out_ch // _BOTTLENECK_EXPANSION
        self.reduce = _ConvBnReLU(in_ch, mid_ch, 1, stride, 0, 1, True)
        self.conv3x3 = _ConvBnReLU(mid_ch, mid_ch, 3, 1, dilation, dilation, True)
        self.increase = _ConvBnReLU(mid_ch, out_ch, 1, 1, 0, 1, False)
        self.shortcut = (
            _ConvBnReLU(in_ch, out_ch, 1, stride, 0, 1, False)
            if downsample
            else lambda x: x  # identity
        )

    def forward(self, x):
        h = self.reduce(x)
        h = self.conv3x3(h)
        h = self.increase(h)
        h += self.shortcut(x)
        return F.relu(h)


class _ResLayer(nn.Sequential):
    """
    Residual layer with multi grids
    """

    def __init__(self, n_layers, in_ch, out_ch, stride, dilation, multi_grids=None):
        super(_ResLayer, self).__init__()

        if multi_grids is None:
            multi_grids = [1 for _ in range(n_layers)]
        else:
            assert n_layers == len(multi_grids)

        # Downsampling is only in the first block
        for i in range(n_layers):
            self.add_module(
                "block{}".format(i + 1),
                _Bottleneck(
                    in_ch=(in_ch if i == 0 else out_ch),
                    out_ch=out_ch,
                    stride=(stride if i == 0 else 1),
                    dilation=dilation * multi_grids[i],
                    downsample=(True if i == 0 else False),
                ),
            )


class _Stem(nn.Sequential):
    """
    The 1st conv layer.
    Note that the max pooling is different from both MSRA and FAIR ResNet.
    """

    def __init__(self, out_ch):
        super(_Stem, self).__init__()
        self.add_module("conv1", _ConvBnReLU(3, out_ch, 7, 2, 3, 1))
        self.add_module("pool", nn.MaxPool2d(3, 2, 1, ceil_mode=True))


if __name__ == "__main__":
    model = DeepLabV3Plus(
        n_classes=21,
        n_blocks=[3, 4, 23, 3],
        atrous_rates=[6, 12, 18],
        multi_grids=[1, 2, 4],
        output_stride=16,
    )
    model.eval()
    image = torch.randn(1, 3, 513, 513)

    print(model)
    print("input:", image.shape)
    print("output:", model(image).shape)

参考链接：


https://blog.csdn.net/junparadox/article/details/52610744



清欢守护者：精读深度学习论文(20) DeepLab V1



https://blog.csdn.net/u011974639/article/details/79518175



语义分割论文-DeepLab系列



语义分割中 CRF 的运用



https://blog.csdn.net/studyeboy/article/details/88121657