mmdetection的ResNet讲解

配置文件中：

 5     backbone=dict(       
  6         type='ResNet',   
  7         depth=101,       
  8         num_stages=4,    
  9         out_indices=(0, 1, 2, 3),
 10         frozen_stages=1,
 11         style='pytorch'),

 

ResNet所在函数：

mmdet/models/backbones/resnet.py

def forward(self, x):
    x = self.conv1(x)
    x = self.norm1(x)
    x = self.relu(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)
    return tuple(outs)

 

因为ResNet的conv1和maxpool本身就会做2次的下采样，所以第0 stage的输出的特征图的stride为8.

self.conv1 = build_conv_layer(
    self.conv_cfg,
    in_channels,
    64,
    kernel_size=7,
    stride=2,
    padding=3,
    bias=False)