感受野计算

感受野

• The receptive field is defined as the region in the input space that a particular CNN’s feature is looking at (i.e. be affected by)
• 感受野被定义为，CNN的某个特征查看输入空间中的区域（即受影响的区域）

感受野计算

• (1) $n_{out}=\lfloor {n_{in}+2p-k \over s} \rfloor + 1$nout​=⌊snin​+2p−k​⌋+1
• $n_{in}$nin​ : number of input features
• $n_{out}$nout​: number of out features
• k: convolution kernel size
• p: convolution padding size
• s: convolution stride size
• (2) $j_{out}=j_{in} * s$jout​=jin​∗s
• (3) $r_{out}=r_{in}+(k-1) * j_{in}$rout​=rin​+(k−1)∗jin​
• (4) $start_{out}=start_{in} + ({k-1\over2}) * j_{in}$startout​=startin​+(2k−1​)∗jin​

对于初始输入图片，jump = 1，等式3计算一个输出特征的感受野大小（size of receptive field，r）

等式4计算第一个输出特征感受野的中心位置，start 是一像素中心坐标（The fourth equation calculates the center position of the receptive field of the first output feature. Here, start is the center coordinate of one pixel.）

• {‘alexnet’: {‘net’:[[11,4,0],[3,2,0],[5,1,2],[3,2,0],[3,1,1],[3,1,1],[3,1,1],[3,2,0]],
  ‘name’:[‘conv1’,‘pool1’,‘conv2’,‘pool2’,‘conv3’,‘conv4’,‘conv5’,‘pool5’]},
• ‘vgg16’: {‘net’:[[3,1,1],[3,1,1],[2,2,0],[3,1,1],[3,1,1],[2,2,0],[3,1,1],[3,1,1],[3,1,1],
  [2,2,0],[3,1,1],[3,1,1],[3,1,1],[2,2,0],[3,1,1],[3,1,1],[3,1,1],[2,2,0]],
  ‘name’:[‘conv1_1’,‘conv1_2’,‘pool1’,‘conv2_1’,‘conv2_2’,‘pool2’,‘conv3_1’,‘conv3_2’,‘conv3_3’,‘pool3’,‘conv4_1’,‘conv4_2’,‘conv4_3’,‘pool4’,‘conv5_1’,‘conv5_2’,‘conv5_3’,‘pool5’]}
• ‘zf-5’:{‘net’: [[7,2,3],[3,2,1],[5,2,2],[3,2,1],[3,1,1],[3,1,1],[3,1,1]],
  ‘name’: [‘conv1’,‘pool1’,‘conv2’,‘pool2’,‘conv3’,‘conv4’,‘conv5’]}}

[1]: https://syncedreview.com/2017/05/11/a-guide-to-receptive-field-arithmetic-for-convolutional-neural-networks/