root_resnet.py

import torch.nn as nn
from mmcv.cnn import xavier_init, constant_init, normal_init
import torch

def conv3x3(in_planes, out_planes, stride=1, dilation=1):
    "3x3 convolutioin with padding"
    return nn.Conv2d(
        in_planes,
        out_planes,
        kernel_size=3,
        stride=stride,
        padding=dilation,
        dilation=dilation,
        bias=False)


class BasicBlock(nn.Module):
    expansion = 1

    def __init__(self,
                 inplanes,
                 planes,
                 stride=1,
                 dilation=1,
                 downsample=None,
                 with_cp=False):
        super(BasicBlock, self).__init__()
        self.conv1 = conv3x3(inplanes, planes, stride, dilation)
        self.bn1 = nn.BatchNorm2d(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = nn.BatchNorm2d(planes)
        self.downsample = downsample
        self.stride = stride
        self.dilation = dilation
        assert not with_cp

    def forward(self, x):
        identity = x

        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)

        out = self.conv2(out)
        out = self.bn2(out)

        if self.downsample is not None:
            identity = self.downsample(x)

        out += identity
        out = self.relu(out)
        return out


class Bottleneck(nn.Module):
    expansion = 4

    def __init__(self,
                 inplanes,
                 planes,
                 stride=1,
                 dilation=1,
                 downsample=None,
                 with_cp=False):
        """Bottleneck block for resnet."""
        super(Bottleneck, self).__init__()
        self.inplanes = inplanes
        self.planes = planes
        self.conv1_stride = 1
        self.conv2_stride = stride

        # self.conv1 doesn't change the size(h, w) of images
        self.conv1 = nn.Conv2d(
            inplanes,
            planes,
            kernel_size=1,
            stride=self.conv1_stride,
            bias=False)

        # self.conv2 will change the size(h, w) of images
        # only if the self.conv2_stride != 1
        self.conv2 = nn.Conv2d(
            planes,
            planes,
            kernel_size=3,
            stride=self.conv2_stride,
            padding=dilation,
            dilation=dilation,
            bias=False)

        self.bn1 = nn.BatchNorm2d(planes)
        self.bn2 = nn.BatchNorm2d(planes)
        # self.conv3 changes the channel of input,
        self.conv3 = nn.Conv2d(
            planes, planes * self.expansion, kernel_size=1, bias=False)
        self.bn3 = nn.BatchNorm2d(planes * self.expansion)
        self.relu = nn.ReLU(inplace=True)
        self.downsample = downsample
        self.stride = stride
        self.dilation = dilation
        self.with_cp = with_cp

    def forward(self, x):

        def _inner_forward(x):
            identity = x

            out = self.conv1(x)
            out = self.bn1(out)
            out = self.relu(out)

            out = self.conv2(out)
            out = self.bn2(out)
            out = self.relu(out)

            out = self.conv3(out)
            out = self.bn3(out)

            if self.downsample is not None:
                identity = self.downsample(identity)

            out += identity

            return out

        if self.with_cp and x.requires_grad:
            out = cp.checkpoint(_inner_forward, x)
        else:
            out = _inner_forward(x)

        out = self.relu(out)

        return out


def make_res_layer(block,
                   inplanes,
                   planes,
                   blocks,
                   stride=1,
                   dilation=1,
                   with_cp=False):
    downsample = None
    if stride != 1 or inplanes != planes * block.expansion:
        downsample = nn.Sequential(
            nn.Conv2d(
                inplanes,
                planes * block.expansion,
                kernel_size=1,
                stride=stride,
                bias=False),
            nn.BatchNorm2d(planes * block.expansion)
        )

    layers = []
    layers.append(
        block(
            inplanes,
            planes,
            stride,
            dilation,
            downsample,
            with_cp=with_cp))
    inplanes = planes * block.expansion
    for i in range(1, blocks):
        layers.append(
            block(inplanes, planes, 1, dilation, with_cp=with_cp))

    return nn.Sequential(*layers)


class SSDRES512(nn.Module):
    arch_setting = {
        18: (BasicBlock, (2, 2, 2, 2)),
        34: (BasicBlock, (3, 4, 6, 3, 1, 1, 1, 1)),
        50: (Bottleneck, (3, 4, 6, 3)),
        101: (Bottleneck, (3, 4, 23, 3)),
        152: (Bottleneck, (3, 8, 36, 3))
    }

    def __init__(self,
                 input_size,
                 depth,
                 num_stages=8,
                 strides=(1, 2, 2, 2, 2, 2, 2, 2),
                 dilations=(1, 1, 1, 1, 1, 1, 1, 1),
                 out_indices=(2, 3, 4, 5, 6, 7),
                 with_cp=False):
        super(SSDRES512, self).__init__()
        self.depth = depth
        self.num_stages = num_stages
        self.strides = strides
        self.dilations = dilations
        self.out_indices = out_indices
        self.with_cp = with_cp

        # stage_blocks = (3, 4, 23, 3)
        self.block, stage_blocks = self.arch_setting[depth]
        self.stage_blocks = stage_blocks[:num_stages]  # (3, 4, 23, 3)
        self.inplanes = 64
        self.input_size = input_size

        self.conv1 = nn.Conv2d(3, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn1 = nn.BatchNorm2d(64)
        self.relu1 = nn.ReLU(inplace=True)
        self.conv2 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn2 = nn.BatchNorm2d(64)
        self.relu2 = nn.ReLU(inplace=True)
        self.conv3 = nn.Conv2d(64, 64, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn3 = nn.BatchNorm2d(64)
        self.relu3 = nn.ReLU(inplace=True)

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

        self.res_layers = []

        for i, num_blocks in enumerate(self.stage_blocks):  # (3, 4, 23, 3)
            stride = strides[i]  # (1, 2, 2, 2)
            dilation = dilations[i]  # (1, 1, 1, 1)
            if i <= 3:
                planes = 64 * 2 ** i
            else:
                planes = 128
            # Bottleneck: input_channels=in_plabnes, output_channels=palnes*4
            res_layer = make_res_layer(
                self.block,
                self.inplanes,
                planes,
                num_blocks,
                stride=stride,
                dilation=dilation,
                with_cp=with_cp)

            self.inplanes = planes * self.block.expansion
            layer_name = 'layer{}'.format(i + 1)
            self.add_module(layer_name, res_layer)
            self.res_layers.append(layer_name)

        self.L2Norm1 = L2Norm(1024, 20)
        self.L2Norm2 = L2Norm(2048, 20)

        self.conv4 = nn.Conv2d(2048, 1024, kernel_size=3, stride=1, padding=1, bias=False)
        self.bn4 = nn.BatchNorm2d(1024)
        self.relu4 = nn.ReLU(inplace=True)
        self.conv5 = nn.Conv2d(1024, 512, kernel_size=3, stride=2, padding=1, bias=False)
        self.bn5 = nn.BatchNorm2d(512)
        self.relu5 = nn.ReLU(inplace=True)
        self.conv6 = nn.Conv2d(512, 256, kernel_size=3, stride=2, padding=1, bias=False)
        self.bn6 = nn.BatchNorm2d(256)
        self.relu6 = nn.ReLU(inplace=True)
        self.conv7 = nn.Conv2d(256, 256, kernel_size=3, stride=1,bias=False)
        self.bn7 = nn.BatchNorm2d(256)
        self.relu7 = nn.ReLU(inplace=True)
        self.conv8 = nn.Conv2d(256, 256, kernel_size=3, stride=1, bias=False)
        self.bn8 = nn.BatchNorm2d(256)
        self.relu8 = nn.ReLU(inplace=True)

    def init_weight(self, pretraind=None):
        if pretraind is None:
            for m in self.modules():
                if isinstance(m, nn.Conv2d):
                    xavier_init(m, distribution='uniform')
                elif isinstance(m, nn.BatchNorm2d):
                    constant_init(m, 1)
                elif isinstance(m, nn.Linear):
                    normal_init(m, std=0.01)
        else:
            raise TypeError('pertrained must be a str or NOne')

    def forward(self, x):
        x = self.conv1(x)
        x = self.bn1(x)
        x = self.relu1(x)
        x = self.conv2(x)
        x = self.bn2(x)
        x = self.relu2(x)
        x = self.conv3(x)
        x = self.bn3(x)
        x = self.relu3(x)
        x = self.maxpool(x)
        outs = []

        for i, layer_name in enumerate(self.res_layers):
            res_layer = getattr(self, layer_name)
            x = res_layer(x)
            if i in self.out_indices:
                outs.append(x)

        outs[0] = self.L2Norm1(outs[0])
        outs[1] = self.L2Norm2(outs[1])

        x = self.conv4(x)
        x = self.bn4(x)
        x = self.relu4(x)
        x = self.conv5(x)
        x = self.bn5(x)
        x = self.relu5(x)
        outs.append(x)
        x = self.conv6(x)
        x = self.bn6(x)
        x = self.relu6(x)
        outs.append(x)
        x = self.conv7(x)
        x = self.bn7(x)
        x = self.relu7(x)
        outs.append(x)
        x = self.conv8(x)
        x = self.bn8(x)
        x = self.relu8(x)
        outs.append(x)

        if len(outs) == 1:
            return outs[0]
        else:
            return tuple(outs)

class L2Norm(nn.Module):
    def __init__(self, in_channels, scale):
        super(L2Norm, self).__init__()
        self.in_channels = in_channels
        self.gamma = scale or None
        self.eps = 1e-10
        self.weight = nn.Parameter(torch.Tensor(self.in_channels))
        self.reset_parameters()

    def reset_parameters(self):
        torch.nn.init.constant_(self.weight, self.gamma)

    def forward(self, x):
        norm = x.pow(2).sum(dim=1, keepdim=True).sqrt()+self.eps
        x = torch.div(x, norm)
        out = self.weight.unsqueeze(0).unsqueeze(2).unsqueeze(3 ).expand_as(x) * x
        return out


if __name__ == "__main__":
    import torch

    input = torch.randn(size=(1, 3, 512, 512))
    net = SSDRES512(input_size=(512, 512), depth=101)
    out = net(input)
    x = out
    for x in out:
        print(x.shape)