def preresnet_bottleneck(l, ch_out, stride, preact):
# stride is applied on the second conv, following fb.resnet.torch
l, shortcut = apply_preactivation(l, preact)
l = Conv2D('conv1', l, ch_out, 1, activation=BNReLU)
l = Grconv('conv2', l, ch_out, 3, strides=stride, activation=BNReLU)
l = Conv2D('conv3', l, ch_out * 4, 1)
return l + resnet_shortcut(shortcut, ch_out * 4, stride)
def preresnet_bottleneck(l, ch_out, stride, preact):
# stride is applied on the second conv, following fb.resnet.torch
l, shortcut = apply_preactivation(l, preact) # 先预激活
l = Conv2D('conv1', l, ch_out, 1, activation=BNReLU)
l = Grconv('conv2', l, ch_out, 3, strides=stride, activation=BNReLU)
l = Conv2D('conv3', l, ch_out * 4, 1)
# l是F(x),shortcut是x！！！！！！！！！！！！！！！！
return l + resnet_shortcut(
shortcut, ch_out * 4,
stride) # l 和shortcut可以相加的条件是他们的维度和通道数相同，让shortcut和ch*4相同
def resnet_bottleneck(l, ch_out, stride, stride_first=False):
"""
stride_first: original resnet put stride on first conv. fb.resnet.torch put stride on second conv.
"""
shortcut = l
l = Conv2D('conv1',
l,
ch_out,
1,
strides=stride if stride_first else 1,
activation=BNReLU)
l = Grconv('conv2',
l,
ch_out,
3,
strides=1 if stride_first else stride,
activation=BNReLU)
l = Conv2D('conv3', l, ch_out * 4, 1, activation=get_bn(zero_init=True))
return l + resnet_shortcut(
shortcut, ch_out * 4, stride, activation=get_bn(zero_init=False))
def preresnet_basicblock(l, ch_out, stride, preact):
l, shortcut = apply_preactivation(l, preact) # 先预激活
# l是输入 ch_out是卷积核，3是kernel size
l = Grconv('conv1', l, ch_out, 3, strides=stride, activation=BNReLU)
l = Grconv('conv2', l, ch_out, 3)
return l + resnet_shortcut(shortcut, ch_out, stride)
def resnet_basicblock(l, ch_out, stride):
shortcut = l
l = Grconv('conv1', l, ch_out, 3, strides=stride, activation=BNReLU)
l = Grconv('conv2', l, ch_out, 3, activation=get_bn(zero_init=True))
return l + resnet_shortcut(
shortcut, ch_out, stride, activation=get_bn(zero_init=False))