一文搞懂深度学习中的感受野（receptive field）

一文搞懂深度学习中的感受野（receptive field）

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).
—— A guide to receptive field arithmetic for Convolutional Neural Networks

1. 概念

感受野（receptive field）是深度学习中一个特别重要的概念，是指经过卷积操作后的特征图对应到输入图像的区域。如图所示，蓝色区域为输入的图像，经过33卷积，stride为1，padding为1，后得到 33的绿色区域（计算公式为

（out = in - kernel_size + 2padding）/ stride + 1 ），再经过33的卷积(stride 为1，padding 为1)后，得到橙色的22区域，那么对于橙色特征图的其中一个特征，其对应到输入图像的感受野就是55。

2. 推算

为了更清楚的解释感受野，下面对图像进行编码，图像是由一个一个的像素组成，如下图

接下来我们使用一种并不常见的方式来展示CNN的层与层之间的关系（如下图，请将脑袋向左倒45°观看>_<），并且配上我们对原图像的编号。

图中黑色的数字表示原图像，蓝色表示卷积操作。用$r_n$来表示第 $n$个卷积层中，每个单元的感受野（即数字序列的长度）；用 $k_n$ 和 $s_n$分别表示第 $n$个卷积层的kernel_size和stride。序列[1 2 3]表示fmap1的第一个单元能看见原图像中的1，2，3这三个像素，而第二个单元则能看见3，4，5。这两个单元随后又被kernel_size=2，stride 1的Filter 2进行卷积，因而得到的fmap2的第一个单元能够看见原图像中的1,2,3,4,5共5个像素（即取[1 2 3]和[3 4 5]的并集）。

接下来我们尝试一下如何用公式来表述上述过程。可以看到，[1 2 3]和[3 4 5]之间因为Filter 1的stride 2而错开（偏移）了两位，而3是重叠的。对于卷积两个感受野为3的上层单元，下一层最大能获得的感受野为 3 + 2 * 1 = 5 。
$r_2$ = $r_1$ + ($k_2-1)$ *$s_1$
继续往下一层看，我们会发现[1 2 3 4 5]和[3 4 5 6 7]的偏移量仍为2，并不简单地等于上一层的
，这是因为之前的stride对后续层的影响是永久性的，而且是累积相乘的关系（例如，在fmap3中，偏移量已经累积到4了），也就是说
$r_3$ = $r_2$ + ( $k_3-1$ ) * $s_2$ * $s_1$,
以此类推，
$r_4$ = $r_3$ + ( $k_4-1$ ) * $s_3$ * $s_2$ * $s_1$

得出递推公式
$r_n$ = $r_{n-1}$ + ( $k_n-1$ ) * ${\prod }_{i=1}^{n-1}{s}_i$

3. 补充

4.代码

接下来，给出代码

computeReceptiveField.py

# [filter size, stride, padding]
#Assume the two dimensions are the same
#Each kernel requires the following parameters:
# - k_i: kernel size
# - s_i: stride
# - p_i: padding (if padding is uneven, right padding will higher than left padding; "SAME" option in tensorflow)
# 
#Each layer i requires the following parameters to be fully represented: 
# - n_i: number of feature (data layer has n_1 = imagesize )
# - j_i: distance (projected to image pixel distance) between center of two adjacent features
# - r_i: receptive field of a feature in layer i
# - start_i: position of the first feature's receptive field in layer i (idx start from 0, negative means the center fall into padding)

import math
convnet =   [[11,4,0],[3,2,0],[5,1,2],[3,2,0],[3,1,1],[3,1,1],[3,1,1],[3,2,0],[6,1,0], [1, 1, 0]]
layer_names = ['conv1','pool1','conv2','pool2','conv3','conv4','conv5','pool5','fc6-conv', 'fc7-conv']
imsize = 227

def outFromIn(conv, layerIn):
  n_in = layerIn[0]
  j_in = layerIn[1]
  r_in = layerIn[2]
  start_in = layerIn[3]
  k = conv[0]
  s = conv[1]
  p = conv[2]
  
  n_out = math.floor((n_in - k + 2*p)/s) + 1
  actualP = (n_out-1)*s - n_in + k 
  pR = math.ceil(actualP/2)
  pL = math.floor(actualP/2)
  
  j_out = j_in * s
  r_out = r_in + (k - 1)*j_in
  start_out = start_in + ((k-1)/2 - pL)*j_in
  return n_out, j_out, r_out, start_out
  
def printLayer(layer, layer_name):
  print(layer_name + ":")
  print("\t n features: %s \n \t jump: %s \n \t receptive size: %s \t start: %s " % (layer[0], layer[1], layer[2], layer[3]))
 
layerInfos = []
if __name__ == '__main__':
#first layer is the data layer (image) with n_0 = image size; j_0 = 1; r_0 = 1; and start_0 = 0.5
  print ("-------Net summary------")
  currentLayer = [imsize, 1, 1, 0.5]
  printLayer(currentLayer, "input image")
  for i in range(len(convnet)):
    currentLayer = outFromIn(convnet[i], currentLayer)
    layerInfos.append(currentLayer)
    printLayer(currentLayer, layer_names[i])
  print ("------------------------")
  layer_name = raw_input ("Layer name where the feature in: ")
  layer_idx = layer_names.index(layer_name)
  idx_x = int(raw_input ("index of the feature in x dimension (from 0)"))
  idx_y = int(raw_input ("index of the feature in y dimension (from 0)"))
  
  n = layerInfos[layer_idx][0]
  j = layerInfos[layer_idx][1]
  r = layerInfos[layer_idx][2]
  start = layerInfos[layer_idx][3]
  assert(idx_x < n)
  assert(idx_y < n)
  
  print ("receptive field: (%s, %s)" % (r, r))
  print ("center: (%s, %s)" % (start+idx_x*j, start+idx_y*j))

5.例子

AlexNet的例子

参考文献：
【1】A guide to receptive field arithmetic for Convolutional Neural Networks
【2】深度神经网络中的感受野(Receptive Field)
【3】彻底搞懂感受野的含义与计算
【4】computeReceptiveField.py