amdegroot/ssd.pytorch 代码来源
vgg基础网络
网络的backbone为vgg，构建vgg网络代码如下，输入为vgg各卷积层通道数和池化层参数cfg，输入图像通道数i，
上图中的Conv6和Conv7对应于Conv6和Conv7两个19*19*1024的特征图，其中Conv7用于预测
我不得不说，这个代码写得很棒
cfg = [64, 64, 'M', 128, 128, 'M', 256, 256, 256, 'C', 512, 512, 512, 'M', 512, 512, 512] def vgg(cfg, i, batch_norm=False): layers = [] in_channels = i for v in cfg: if v == 'M': layers += [nn.MaxPol2d(kernel_size=2, stride=2)] elif v == 'C': layers += [nn.MaxPol2d(kernel_size=2, stride=2, ceil_mode=True)] else: conv2d = nn.Conv2d(in_channels, v, kernel_size=3, padding=1) if batch_norm: layers += [conv2d, nn.Batchnorm2d(v), nn.ReLU(inplace=True)] else: layers += [conv2d, nn.ReLU(inplace=True)] in_channels = v pool5 = nn.MaxPol2d(kernel_size=3, stride=1, padding=1) conv6 = nn.Conv2d(512, 1024, kernel_size=3, padding=6, dilation=6) conv7 = nn.Conv2d(1024, 1024, kernel_size=1) layers += [pool5， conv6, nn.ReLU(inplace=True), conv7, nn.ReLU(inplace=True)] return layers
二、额外的卷积层
SSD在基础vgg网络后面添加了额外的卷积层，通过这些卷积层获得的feature 19*19、10*10、5*5、3*3、1*1等map逐级减少，
通过预测这些层，可以获得多尺度的效果。构建代码如下，因为Conv6和Conv7是在之前的vgg函数中构建的，
这里的输入是Conv7获得的19*19*1024feature map。cfg是要构建的卷积层的通道数，s表示这一层需要stride=2
cfg = [256, 'S', 512, 128, 'S', 256, 128, 256, 128, 256] def add_extras(cfg, i, batch_norm=False): # Extra layers added to VGG for feature scaling # 1*1和3*3的卷积交替交替 layers = [] in_channels = i flag = False for k, v in enumerate(cfg): if in_channels != 'S': if v
== 'S': layers += [nn.Conv2d(in_channels, cfg[k + 1], kernel_size=(1, 3)[flag], stride=2, padding=1)] else: layers += [nn.Conv2d(in_channels, v, kernel_size=(1, 3)[flag])] flag = not flag in_channels = v return layers
三、multibox层
这一层是与图1中的classifier相对应的分类和位置回归层。loc_layers由6个输出维度default_box_num * 4.3*3卷积层组成，
conf_layers由6个输出维度default_box_num * class_num的3*3卷积层 cfg = [4, 6, 6, 6, 4, 4] #每个预测层的defaultt box个数 def multibox(vgg, extra_layers, cfg, num_classes):
loc_layers = [] #为loc回归层和conf预测层创建两个listt conf_layers = [] #在vgglayer中，索引为21和-2的卷积输出feature map对应于38*38*512和19*19*1024的预测层 vgg_source = [21, -2] for k, v in enumerate(vgg_source): loc_layers += [nn.Conv2d(vgg[v].out_channels, cfg[k] * 4, kernel_size=3, padding=1)] conf_layers += [nn.Conv2d(vgg[v].out_channels, cfg[k] * num_classes, kernel_size=3, padding=1)] for k, v in enumerate(extra_layers[1::2], 2): #k从2开始 #每两个索引在extra_layerslist中获取一次，即取3*3conv层 loc_layers += [nn.Conv2d(v.out_channels, cfg[k] * 4, kernel_size=3, padding=1)] conf_layers += [nn.Conv2d(v.out_channels, cfg[k] * num_classes, kernel_size=3, padding=1)] return vgg, extra_layers, (loc_layers, conf_layers)
四、default_box的生成
这部分代码在prior_box.PriorBox在py文件中。
根据不同的scale和ratio，对每个特征图生成kdefaultt boxes
def forward(self): """ 测试中的参数如下 feature_maps: <class 'list'>: [38, 19, 10, 5, 3, 1] steps: <class 'list'>: [8, 16, 32, 64, 100, 300] min_size: <class 'list'>: [30, 60, 111, 162, 213, 264] max_size: <class 'list'>: [60, 111, 162, 213, 264, 315] aspect_ratios: <class 'list'>: [[2], [2, 3], [2, 3], [2, 3], [2], [2]] """ mean = [] #6个特征图遍历 for k, f in enumerate(self.feature_maps): #遍历特征图中的每一个点 for i, j in product(range(f), repeat=2): f_k = self.image_size / self.steps[k] #特征图一点对应原图的大小 # unit center x,y #以上默认框中心设置的公式计算默认框中心，将每一点归一化 cx = (j + 0.5) / f_k cy = (i + 0.5) / f_k # aspect_ratio: 1、长宽比为1，以min_size为边长的默认框架 # rel size: min_size s_k = self.min_sizes[k]/self.image_size mean += [cx, cy, s_k, s_k] # aspect_ratio: 1.当长宽比为1时，增加一个额外的尺度 # rel size: 计算公式sqrt(s_k * s_(k+1))，s_k+等于这一层的max__size s_k_prime = sqrt(s_k * (self.max_sizes[k]/self.image_size)) mean += [cx, cy, s_k_prime, s_k_prime] # rest of aspect ratios，计算其他长宽比的默认框，按公式分别遵循w和h for ar in self.aspect_ratios[k]: mean += [cx, cy, s_k*sqrt(ar), s_k/sqrt(ar)] mean += [cx, cy, s_k/sqrt(ar), s_k*sqrt(ar)] # back to torch land，output是(8732，4)Tensor output = torch.Tensor(mean).view(-1, 4) if self.clip: output.clamp_(max=1, min=0) #output的控制范围为[0，1] return output
湘公网安备43019002002060