【语义分割】ASPP

ASPP

空洞空间卷积池化金字塔(atrous spatial pyramid pooling (ASPP))对所给定的输入以不同采样率的空洞卷积并行采样，相当于以多个比例捕捉图像的上下文。

上图为deeplab v2的ASPP模块，deeplabv3中向ASPP中添加了BN层，其中空洞卷积的rate的意思是在普通卷积的基础上，相邻权重之间的间隔为rate-1, 普通卷积的rate默认为1，所以空洞卷积的实际大小为$k+(k-1)(rate-1)$，其中k为原始卷积核大小。

输出大小如何计算？

问题：当rate接近feature map大小时，$3\times 3$滤波器不是捕获全图像上下文，而是退化为简单的$1\times 1$滤波器，只有滤波器中心起作用。

改进：Concat（ $1\times 1$卷积 ， 3个 $3\times 3$空洞卷积 +，pooled image feature）并且每个卷积核都有256个且都有BN层。

#without bn version
class ASPP(nn.Module):
    def __init__(self, in_channel=512, depth=256):
        super(ASPP,self).__init__()
        self.mean = nn.AdaptiveAvgPool2d((1, 1)) #(1,1)means ouput_dim
        self.conv = nn.Conv2d(in_channel, depth, 1, 1)
        self.atrous_block1 = nn.Conv2d(in_channel, depth, 1, 1)
        self.atrous_block6 = nn.Conv2d(in_channel, depth, 3, 1, padding=6, dilation=6)
        self.atrous_block12 = nn.Conv2d(in_channel, depth, 3, 1, padding=12, dilation=12)
        self.atrous_block18 = nn.Conv2d(in_channel, depth, 3, 1, padding=18, dilation=18)
        self.conv_1x1_output = nn.Conv2d(depth * 5, depth, 1, 1)
 
    def forward(self, x):
        size = x.shape[2:]
 
        image_features = self.mean(x)
        image_features = self.conv(image_features)
        image_features = F.upsample(image_features, size=size, mode='bilinear')
 
        atrous_block1 = self.atrous_block1(x)
        atrous_block6 = self.atrous_block6(x)
        atrous_block12 = self.atrous_block12(x)
        atrous_block18 = self.atrous_block18(x)
 
        net = self.conv_1x1_output(torch.cat([image_features, atrous_block1, atrous_block6,
                                              atrous_block12, atrous_block18], dim=1))
        return net