ASPP
DeepLab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected CRFs.论文中提出的一种可提高感受野的技术。
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 ) ( r a t e − 1 ) k+(k-1)(rate-1)k+(k−1)(rate−1)，其中k为原始卷积核大小。
问题：当rate接近feature map大小时，3 × 3 滤波器不是捕获全图像上下文，而是退化为简单的1 × 1 滤波器，只有滤波器中心起作用。
解决方案：Concat（ 1 × 1 卷积 ， 3个 3 × 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 versionclass 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
转载：https://blog.csdn.net/qq_36530992/article/details/102628455