Tensorflow——理解tf.nn.conv2d方法

作者: SpareNoEfforts | 来源:发表于2018-10-30 10:29 被阅读22次
定义
tf.nn.conv2d (input, filter, strides, padding, use_cudnn_on_gpu=None, data_format=None, name=None)
input: 输入的要做卷积的图片，要求为一个张量，shape为 [ batch, in_height, in_weight, in_channel ]，其中batch为图片的数量，in_height 为图片高度，in_weight 为图片宽度，in_channel 为图片的通道数，灰度图该值为1，彩色图为3。
filter：卷积核，要求也是一个张量，shape为 [ filter_height, filter_weight, in_channel, out_channels ]，其中 filter_height 为卷积核高度，filter_weight 为卷积核宽度，in_channel 是图像通道数，和 input 的 in_channel 要保持一致，out_channel 是卷积核数量。
stride:卷积时在图像每一维的步长，这是一个一维的向量，[ 1, strides, strides, 1]，第一位和最后一位固定必须是1
padding:string类型，值为“SAME” 和 “VALID”，表示的是卷积的形式，是否考虑边界。"SAME"是考虑边界，不足的时候用0去填充周围，"VALID"则不考虑
即为：
SAME：卷积核可以停留在图像边缘，即，宽度为Image_weight / stride + 1
VALID : 卷积核不可以停留在图像边缘，宽度为(Image_weight - filter_weight)/stride +1
use_cudnn_on_gpu： bool类型，是否使用cudnn加速，默认为true
举例
import tensorflow as tf
------------------------------------------------------------------------------------------------------
# case 1
# 输入是1张 3*3 大小的图片，图像通道数是5，卷积核是 1*1 大小，数量是1
# 步长是[1,1,1,1]最后得到一个 3*3 的feature map
# 1张图最后输出就是一个 shape为[1,3,3,1] 的张量
input = tf.Variable(tf.random_normal([1,3,3,5]))
filter = tf.Variable(tf.random_normal([1,1,5,1]))
op1 = tf.nn.conv2d(input, filter, strides=[1,1,1,1], padding='SAME')
#输出
******************** op1 ********************
[[[[ 0.78366613]
   [-0.11703026]
   [ 3.533338  ]]

  [[ 3.4455981 ]
   [-2.40102   ]
   [-1.3336506 ]]

  [[ 1.9816184 ]
   [-3.3166158 ]
   [ 2.0968733 ]]]]
-------------------------------------------------------------------------------------------------

# case 2
# 输入是1张 3*3 大小的图片，图像通道数是5，卷积核是 2*2 大小，数量是1
# 步长是[1,1,1,1]最后得到一个 3*3 的feature map
# 1张图最后输出就是一个 shape为[1,3,3,1] 的张量 
input = tf.Variable(tf.random_normal([1,3,3,5]))
filter = tf.Variable(tf.random_normal([2,2,5,1]))
op2 = tf.nn.conv2d(input, filter, strides=[1,1,1,1], padding='SAME')

#输出
******************** op2********************
[[[[ 1.9288185]
   [-3.872772 ]
   [ 1.1486561]]

  [[-1.6453924]
   [ 1.9384439]
   [ 1.5341384]]

  [[-0.6960499]
   [ 3.108924 ]
   [ 0.6097143]]]]
-------------------------------------------------------------------------------------------------

# case 3  
# 输入是1张 3*3 大小的图片，图像通道数是5，卷积核是 1*1 大小，数量是1
# 步长是[1,1,1,1]最后得到一个 3*3 的feature map (不考虑边界)
# 1张图最后输出就是一个 shape为[1,3,3,1] 的张量
input = tf.Variable(tf.random_normal([1,3,3,5]))  
filter = tf.Variable(tf.random_normal([1,1,5,1]))  
op3 = tf.nn.conv2d(input, filter, strides=[1, 1, 1, 1], padding='VALID') 

#输出
******************** op3********************
[[[[ 1.6719851 ]
   [-0.38163936]
   [-0.13288862]]

  [[-0.11872166]
   [-0.05304074]
   [ 1.8199785 ]]

  [[ 1.5945687 ]
   [-0.38205272]
   [ 0.28184414]]]]
------------------------------------------------------------------------------------------------- 
# case 4
# 输入是1张 5*5 大小的图片，图像通道数是5，卷积核是 3*3 大小，数量是1
# 步长是[1,1,1,1]最后得到一个 3*3 的feature map (不考虑边界)
# 1张图最后输出就是一个 shape为[1,3,3,1] 的张量
input = tf.Variable(tf.random_normal([1,5,5,5]))  
filter = tf.Variable(tf.random_normal([3,3,5,1]))  
op4 = tf.nn.conv2d(input, filter, strides=[1, 1, 1, 1], padding='VALID')  

#输出
******************** op4********************
[[[[ 4.93663   ]
   [11.300637  ]
   [-4.7170043 ]]

  [[ 4.4999595 ]
   [ 3.7627645 ]
   [ 7.033349  ]]

  [[ 5.1601    ]
   [ 1.9807271 ]
   [-0.28120422]]]]
-------------------------------------------------------------------------------------------------
# case 5  
# 输入是1张 5*5 大小的图片，图像通道数是5，卷积核是 3*3 大小，数量是1
# 步长是[1,1,1,1]最后得到一个 5*5 的feature map (考虑边界)
# 1张图最后输出就是一个 shape为[1,5,5,1] 的张量
input = tf.Variable(tf.random_normal([1,5,5,5]))  
filter = tf.Variable(tf.random_normal([3,3,5,1]))  
op5 = tf.nn.conv2d(input, filter, strides=[1, 1, 1, 1], padding='SAME')  

#输出
******************** op5********************
[[[[  0.615875  ]
   [  2.6127    ]
   [ -3.2841504 ]
   [  9.918146  ]
   [ -2.5437124 ]]

  [[-11.686163  ]
   [ -2.0402982 ]
   [ -1.4028796 ]
   [ -3.1973681 ]
   [  9.569347  ]]

  [[ -3.488454  ]
   [ 11.750545  ]
   [ -5.4785585 ]
   [ -5.6196046 ]
   [ -5.6271753 ]]

  [[  0.06084216]
   [  4.0636263 ]
   [ -0.11393895]
   [  2.4829671 ]
   [  1.6683083 ]]

  [[  6.3058996 ]
   [  1.3491695 ]
   [ -0.10401154]
   [  5.943573  ]
   [ -2.1095133 ]]]]
-------------------------------------------------------------------------------------------------
# case 6 
# 输入是1张 5*5 大小的图片，图像通道数是5，卷积核是 3*3 大小，数量是7
# 步长是[1,1,1,1]最后得到一个 5*5 的feature map (考虑边界)
# 1张图最后输出就是一个 shape为[1,5,5,7] 的张量
input = tf.Variable(tf.random_normal([1,5,5,5]))  
filter = tf.Variable(tf.random_normal([3,3,5,7]))  
op6 = tf.nn.conv2d(input, filter, strides=[1, 1, 1, 1], padding='SAME')  

#输出
******************** op6********************
[[[[  3.0887563   -8.757811    -1.9566984   -3.3707764    1.4573003
     -2.9470344    4.2769175 ]
   [  6.8355904   -2.8316915    4.151598    -1.9893703    4.5343633
     -1.7455567    2.3923404 ]
   [ -0.5796106   -8.3059025    2.6339238   -0.407861     4.038948
     -0.73221153   5.5349445 ]
   [  1.0038129   -0.23307621   8.721055    -8.745911    -1.0426689
      2.7741075    5.2638793 ]
   [ -3.619012     9.038741     3.6280208    2.6974936    1.7085941
      0.25397038  -3.354182  ]]

  [[  3.2160585   -4.840355     3.6852489  -12.032576    -9.674893
     -6.2895727   -4.815977  ]
   [  1.4725536   -4.2549667   -3.4826493   -8.04364      0.25006047
      7.7094593    0.86553353]
   [ 11.234627     6.4976263   -1.7011687   -1.9686639   -5.7297363
      1.469283     5.1969666 ]
   [ -1.0064669   11.335008     5.2906194    1.5887578    2.5739172
     -6.681587   -10.565671  ]
   [ 12.442774     2.3762116   -0.6160493    6.727649    -0.5381024
      5.1488156    4.6439753 ]]

  [[-15.513361    -2.3790777   -1.9252743    2.2043197   -1.4400063
      0.75842166 -10.506536  ]
   [  4.564708    -0.4744407    7.4912853   14.069539    11.6688595
     -4.612417     5.4330025 ]
   [-11.14007      2.9735382   -0.4313957   -4.6811137  -11.4551
     -3.5916057    0.4944711 ]
   [ -3.8912075   -3.9660525    0.91355586  -1.3563845   -1.2875407
      7.572529    -1.4311553 ]
   [  8.351269     0.19171536  -6.96131      1.5982566    0.5016942
      2.9076457   -0.18807316]]

  [[  4.8925915   -5.543425    -0.68927133  -7.493097    -7.4924517
     -7.3695526   12.963789  ]
   [ 10.933542    11.5189495   -1.674472     1.5411812   14.764674
     -7.282252    -5.4992485 ]
   [  0.05787128  -4.8447866    6.0995793   -2.130572     4.788317
      4.4105816   -0.76676995]
   [  2.4877496   -9.822114     1.5490786  -11.292644    -6.0652895
     -5.461935    10.499823  ]
   [  1.1236672    7.7374783   11.541739     0.2169593   -1.80403
     -0.37783074   7.946163  ]]

  [[ -4.1483784    3.2257776    6.348017    -2.6234756   -0.872122
     -5.1639204   -1.8608298 ]
   [  5.9412355   -5.605309    -7.8225245   -3.0232863   -4.3902245
     -1.4678087   -2.756116  ]
   [ -0.5923271   -0.6401453   -0.51332414  -2.9427621   -4.3485436
      5.7885294    5.46108   ]
   [ -3.3316593    5.2753196    2.0999293  -16.171337     1.4345255
     -0.9159676   -3.196972  ]
   [  7.209021     3.3031735   -0.87703556  -2.6866136    0.5111834
      6.312125     4.067911  ]]]]
-------------------------------------------------------------------------------------------------
# case 7  
# 输入是1张 5*5 大小的图片，图像通道数是5，卷积核是 3*3 大小，数量是7
# 步长是[1,2,2,1]最后得到7个 3*3 的feature map (考虑边界)
# 1张图最后输出就是一个 shape为[1,3,3,7] 的张量
input = tf.Variable(tf.random_normal([1,5,5,5]))  
filter = tf.Variable(tf.random_normal([3,3,5,7]))  
op7 = tf.nn.conv2d(input, filter, strides=[1, 2, 2, 1], padding='SAME')  

#输出
******************** op7********************
[[[[ -7.560372     2.0237      -4.8913746   -0.6458702    1.9221705
     -4.1618495   -5.0541553 ]
   [ -3.027947     6.132785    -1.4402869   -4.5081306   -1.0875907
      5.2696166    4.588485  ]
   [  4.8169804    1.6802535    3.742131     0.92543554   3.933595
      7.609642     2.3418703 ]]

  [[  4.678027   -10.652521     3.671758     1.9559444    4.3873854
      0.69699585   6.387135  ]
   [  4.6600924   -0.90035415   9.562813    -4.2748165    5.088637
     -6.6742654    4.0188    ]
   [ -4.3796535    7.632821    -6.2888837    1.8178465   -2.6003487
      0.8694892    2.5623817 ]]

  [[  2.830203     1.5999559   -3.041421    -6.4916563    0.24484622
      0.8900788   -0.68253803]
   [  0.5634527    0.4465307   -1.8369514    4.977104    -0.6233926
      6.5963316    7.077495  ]
   [ -4.4429035    1.1985601    9.393924     3.1787338    2.5441146
      1.5674452   -2.6049848 ]]]]
-------------------------------------------------------------------------------------------------
# case 8  
# 输入是10 张 5*5 大小的图片，图像通道数是5，卷积核是 3*3 大小，数量是7
# 步长是[1,2,2,1]最后每张图得到7个 3*3 的feature map (考虑边界)
# 10张图最后输出就是一个 shape为[10,3,3,7] 的张量
input = tf.Variable(tf.random_normal([10,5,5,5]))  
filter = tf.Variable(tf.random_normal([3,3,5,7]))  
op8 = tf.nn.conv2d(input, filter, strides=[1, 2, 2, 1], padding='SAME')  
[[[[ 3.81172585e+00 -6.58235610e-01  9.39655185e-01 -1.12831044e+00
    -1.44241846e+00 -1.69457889e+00  4.36120272e+00]
   [ 2.00121582e-01 -5.16375589e+00  1.22973609e+00  9.44681168e+00
    -1.00510824e+00 -2.68596339e+00  5.99930286e+00]
   [ 7.93656158e+00 -5.48158646e+00 -1.03805804e+00 -2.67814904e-01
     1.19332771e+01 -7.91494846e+00  5.40462017e-01]]

  [[ 6.18537784e-01  1.22472506e+01  4.73466825e+00  2.60108924e+00
    -2.27779555e+00 -1.39395344e+00  9.39258754e-01]
   [ 2.11174369e-01 -1.11202860e+01  1.15606651e+01 -2.10358667e+00
    -1.91229045e+00  4.73274112e-01 -8.28661203e-01]
   [-1.73816741e+00 -9.91873860e-01 -1.29899430e+00  3.11782837e+00
    -5.76431704e+00 -4.64563322e+00 -6.76457500e+00]]

  [[-3.78995442e+00 -6.68081284e-01  2.67540932e+00  1.45131636e+01
    -5.45289516e+00 -7.98859119e-01  2.98502207e+00]
   [-4.60981560e+00 -2.43963814e+00 -2.20421195e-01  1.00090513e+01
     3.96379256e+00 -1.85508728e-01 -8.32319832e+00]
   [-5.96280575e+00 -5.27051866e-01 -3.18238068e+00 -5.42228127e+00
    -4.69559002e+00  8.09168816e-02  1.91771030e+00]]]


 [[[ 1.53048587e+00  3.10868233e-01  4.89992142e+00  1.96467185e+00
     6.57781363e-02  4.64759827e+00 -3.45945859e+00]
   [ 7.76488018e+00 -3.91228616e-01 -5.22320509e-01 -1.11246891e+01
     4.06208277e+00 -3.44819212e+00 -7.47523904e-01]
   [ 5.63401508e+00 -4.41431141e+00 -9.00770283e+00 -4.24449444e-01
    -6.22014999e+00  2.03636217e+00 -9.96613204e-01]]

  [[-5.48789918e-01  5.95879555e+00  1.20323877e+01  1.08595169e+00
    -4.32021350e-01 -3.60252470e-01  1.51402450e+00]
   [-2.18106341e+00  1.20968828e+01 -1.47913229e+00 -6.29183245e+00
    -1.08375347e+00  7.86020947e+00 -2.51888847e+00]
   [-3.51060534e+00 -7.77409649e+00 -3.00079918e+00  4.92167711e-01
    -9.36036873e+00  9.77025986e-01 -4.88964081e+00]]

  [[ 7.76729643e-01  1.08929138e+01  5.02641439e-01 -3.69875813e+00
    -2.96155214e+00  8.67065811e+00  7.10426807e-01]
   [ 5.79438925e+00  7.40104389e+00 -1.15843797e+00 -1.13216896e+01
    -8.39748192e+00  3.47570372e+00  7.48564816e+00]
   [-1.20307946e+00 -2.13643742e+00  5.40993786e+00 -1.48298454e+00
    -4.71552181e+00  9.79691207e-01 -9.91011858e-01]]]


 [[[ 2.50824261e+00 -2.84440899e+00  1.71533370e+00  7.51872396e+00
    -1.03872118e+01 -1.40590143e+00  7.93499374e+00]
   [ 7.06769943e-01  4.55543947e+00 -1.12701893e-01 -1.74570131e+00
     7.37167931e+00 -6.60094738e+00  1.18070545e+01]
   [-3.77333736e+00 -6.92615986e+00  1.58949256e+00  8.52469063e+00
    -3.17621326e+00  4.55142117e+00  1.86487818e+00]]

  [[ 2.62441111e+00  1.51057935e+00 -1.69782043e+00 -1.06731892e+01
     9.10656643e+00 -1.90770149e+00 -3.25653982e+00]
   [ 1.25612850e+01  2.38117480e+00 -4.60387325e+00  1.23946321e+00
    -6.48248291e+00  8.82726860e+00  1.96533287e+00]
   [ 5.56728244e-01 -2.70045996e+00  9.12153625e+00 -7.64613342e+00
     7.03843832e+00  1.47829437e+00 -1.13098145e+00]]

  [[-2.29912233e+00  4.13240147e+00  6.55635834e-01  1.01796589e+01
    -2.68077517e+00 -2.44472027e+00  3.18214369e+00]
   [ 7.21115589e+00  3.16968679e+00  1.12844181e+01 -3.52053165e+00
    -4.21286821e-02 -4.63994265e-01  7.78382206e+00]
   [ 3.35062027e-01  1.86070824e+00  4.23208809e+00 -1.28888645e+01
     2.51626110e+00 -7.29913950e-01  4.48049974e+00]]]


 [[[-2.28839159e+00 -1.83935964e+00  4.67500973e+00 -8.66911113e-01
     7.57640362e+00 -4.49133158e+00 -3.58713269e+00]
   [-2.56349277e+00  1.65234530e+00 -3.40921402e+00 -3.57537770e+00
    -1.12090588e+00 -4.00662136e+00 -2.31218624e+00]
   [-1.59960270e+00 -1.02549422e+00  1.70352447e+00 -5.07849455e-02
     5.91472030e-01 -6.75261378e-01 -2.98815727e+00]]

  [[-4.95550632e+00 -1.59211278e+00  2.85315514e+00 -3.60998213e-01
     1.10690346e+01 -2.92121983e+00 -8.75728607e+00]
   [ 1.04558921e+00 -1.10802925e+00  2.20073867e+00  6.86034966e+00
    -8.90627098e+00 -8.71938992e+00 -4.64016318e-01]
   [ 6.93283558e+00 -7.52256966e+00  2.89755774e+00 -1.97585964e+00
     1.34779205e+01 -7.09957218e+00  2.72134781e+00]]

  [[-9.07215500e+00  7.41471624e+00  5.28336763e+00 -4.50483036e+00
     8.20894718e-01  5.68455124e+00 -2.32907462e+00]
   [ 2.75068545e+00  9.57029152e+00 -3.07973957e+00 -3.78625989e+00
    -8.29107571e+00  3.85373855e+00  1.01118898e+00]
   [ 6.30977154e-02  4.62665367e+00 -4.43314362e+00  1.64909089e+00
    -2.61656308e+00  2.18150091e+00 -1.97284305e+00]]]


 [[[-4.68610287e+00  4.63123989e+00  1.93239570e+00 -6.24144125e+00
    -2.12857890e+00  7.83749342e+00  3.71262431e+00]
   [-5.97846508e-01 -4.51939315e-01  4.12002516e+00 -1.59932184e+00
    -1.24146347e+01 -3.81853724e+00  6.19719982e+00]
   [ 5.69620705e+00  1.17846155e+00 -3.76912522e+00 -2.78199720e+00
    -4.14124632e+00 -8.70480537e-01  7.28857422e+00]]

  [[ 6.51047325e+00 -9.75043869e+00 -7.16108227e+00  3.37941360e+00
     1.78760872e+01 -1.79757142e+00 -9.61674118e+00]
   [ 6.94279099e+00 -2.50615239e+00  5.16652727e+00 -9.33402729e+00
     7.53117752e+00 -4.19789219e+00 -1.47018957e+00]
   [ 6.32885885e+00  3.85915112e+00  5.42673349e-01 -5.80658555e-01
    -1.97605848e-01 -1.72271812e+00 -5.95912266e+00]]

  [[-3.86114836e+00  1.97221160e-01 -3.30170727e+00 -3.46406794e+00
     3.63580871e+00  9.82056856e-01 -1.79573965e+00]
   [-4.00891447e+00  8.22873688e+00  8.06672668e+00 -4.28653002e+00
    -5.32858133e+00  9.43519115e+00 -3.52467918e+00]
   [-4.78855515e+00  5.19511986e+00 -4.68266487e+00  1.17632685e+01
    -4.77883863e+00 -1.88284159e-01 -5.56722975e+00]]]


 [[[ 1.89772367e-01  3.16071630e-01  3.13395715e+00  3.01754177e-01
     4.14337683e+00 -6.85967159e+00  2.19734955e+00]
   [-9.88918841e-01  3.14887810e+00 -3.53439212e+00 -2.68401289e+00
    -4.27296972e+00  3.57035112e+00  4.22077084e+00]
   [ 3.00699210e+00  3.36317134e+00 -9.01305914e-01 -8.24847412e+00
     5.16212940e-01  4.10071039e+00  8.48048019e+00]]

  [[ 5.30344546e-01  8.90641308e+00 -5.22371292e+00 -9.62609529e-01
    -8.23948574e+00  1.11528502e+01  4.19720745e+00]
   [ 5.41305363e-01  7.66341686e+00  8.16087127e-01 -1.69427891e+01
    -9.43431139e-01  3.76207352e+00  2.09374666e+00]
   [-1.97999740e+00 -1.78795707e+00  1.24396682e+00 -5.56382322e+00
    -7.98738337e+00  5.25424480e+00 -2.88212180e+00]]

  [[ 1.11028862e+01 -9.86741066e+00 -2.52228451e+00  1.31483240e+01
    -1.80064118e+00 -4.19618654e+00  4.12672377e+00]
   [-2.38205004e+00  3.45358086e+00  3.06554794e+00  1.39153457e+00
     1.74696178e+01 -9.50703239e+00 -7.38960409e+00]
   [ 1.88154280e-02  6.71151447e+00  7.40723515e+00 -6.85136318e-01
    -2.66253185e+00  1.28768826e+00 -1.88888133e+00]]]


 [[[-1.81933379e+00  7.67806411e-01 -3.73196530e+00  1.59054911e+00
    -3.48422110e-01  9.18928087e-01 -5.09416533e+00]
   [-7.46384335e+00  9.51658821e+00  2.05104017e+00  4.65793419e+00
    -7.38202190e+00 -3.89330769e+00  1.22091656e+01]
   [-6.46189404e+00 -2.52824783e-01 -2.53815079e+00  2.52569151e+00
     2.53525639e+00  1.75139821e+00  1.39415288e+00]]

  [[-5.36459589e+00 -7.83692455e+00  4.92164993e+00 -1.64086747e+00
     3.42228937e+00 -1.28203106e+00 -6.44423771e+00]
   [ 6.13428068e+00  2.02679372e+00 -8.31424177e-01 -1.11501341e+01
    -6.76840162e+00  8.00881684e-01  8.87121487e+00]
   [-3.81009316e+00 -1.20232134e+01  1.93391097e+00  4.11974144e+00
    -1.14694452e+00  3.36693192e+00 -3.51179051e+00]]

  [[-4.13361931e+00  2.33749223e+00  1.10010195e+01 -3.48916125e+00
     9.83463669e+00  1.14318848e-01 -1.02951355e+01]
   [-2.07277107e+00  4.58487415e+00 -2.85462832e+00  9.62093830e-01
    -1.10009775e+01  2.91307449e-01 -2.71501541e-02]
   [ 1.35933065e+00  2.23298883e+00 -1.53114724e+00 -2.99561787e+00
     5.05122375e+00 -4.47726011e-01 -1.30996180e+00]]]


 [[[-2.30837297e+00  2.31148410e+00  1.62851739e+00 -7.04085350e+00
     1.48006344e+00 -8.36053848e-01 -2.05071878e+00]
   [-1.38948321e+00 -1.79432559e+00  1.66157138e+00  2.52863026e+00
    -7.40308714e+00 -1.66243482e+00 -4.31997442e+00]
   [ 4.34525728e-01 -4.86747551e+00  2.63057780e+00 -7.23891139e-01
     1.20453250e+00 -3.09287119e+00  1.81485105e+00]]

  [[-4.34996891e+00  6.45046651e-01  2.17182612e+00  3.30334997e+00
     4.39480066e+00  2.21359074e-01 -9.40736866e+00]
   [ 1.32711506e+00  1.82248173e+01 -3.49929571e+00 -7.32633162e+00
    -6.05510807e+00  4.97397137e+00 -6.42782402e+00]
   [-4.16941261e+00  4.90973854e+00  7.69643211e+00 -2.56850910e+00
    -1.23103590e+01  3.96808362e+00  3.48330426e+00]]

  [[-1.41771734e+00  3.80652380e+00  6.81313419e+00  3.99951291e+00
    -4.59768105e+00 -2.67537093e+00  3.28954649e+00]
   [ 3.44624519e-02  3.46278954e+00 -7.19838619e-01 -7.18024731e+00
     4.72416639e+00 -8.42012405e+00  1.02093210e+01]
   [ 1.77111924e+00  1.76189995e+00 -4.04496670e+00 -8.58612251e+00
    -4.48928452e+00  1.99987328e+00  2.11980033e+00]]]


 [[[ 5.16344547e+00 -3.28772712e+00  2.55633926e+00 -1.51607990e-02
    -4.33142781e-01 -6.92938566e+00  3.01811099e+00]
   [ 4.50596237e+00 -2.95426488e+00  6.79820490e+00  6.90650177e+00
     1.18538189e+00 -2.41478145e-01  5.52644825e+00]
   [-3.45425940e+00 -1.01008582e+00  2.75584054e+00  4.67782021e+00
     4.02924681e+00 -3.02245331e+00 -2.42830753e+00]]

  [[-1.24309397e+00  2.25454855e+00 -4.99355435e-01 -4.18008947e+00
    -3.40147209e+00  3.88633633e+00 -1.01025581e+00]
   [ 8.36562634e+00  5.56176138e+00 -3.90685773e+00 -1.72424483e+00
     2.73722672e+00  8.00842190e+00 -2.48433948e+00]
   [-4.46336555e+00 -3.72679162e+00 -1.26352954e+00  4.95191813e+00
    -3.23514009e+00  7.82737851e-01 -7.65897846e+00]]

  [[-3.07640934e+00  1.11932468e+00 -2.62503529e+00 -1.19850216e+01
    -8.01890469e+00  1.48621440e+00  3.36904168e+00]
   [ 3.82285118e-01  2.76505685e+00  1.05706239e+00 -7.88955784e+00
    -4.90172768e+00  1.72243714e+00  2.69844866e+00]
   [-9.21259785e+00 -3.33890224e+00  5.46678066e+00 -1.62324953e+00
    -6.74576473e+00  9.93959785e-01  4.94233227e+00]]]


 [[[-4.70427895e+00  3.07578850e+00  1.34436083e+00  6.60757589e+00
     6.15761566e+00  1.20834816e+00 -1.43940473e+00]
   [-2.02920246e+00  9.61333084e+00 -5.87510967e+00  4.66540051e+00
    -2.92267442e-01 -1.75225294e+00  3.57326508e+00]
   [ 1.95737648e+00 -1.88611352e+00 -1.09006286e+00 -4.27253246e+00
    -2.21520901e-01  9.55001116e-02  2.74844027e+00]]

  [[-6.32184601e+00 -5.88555717e+00 -5.94709516e-01 -4.42621136e+00
    -1.81127739e+00  1.04070306e-01  8.58660412e+00]
   [ 8.64363861e+00 -1.60207062e+01 -1.36644773e+01  1.04534721e+01
     2.78887987e+00  9.04768276e+00 -7.40338182e+00]
   [ 7.05552197e+00 -4.00396538e+00  3.59226894e+00 -5.75797653e+00
     7.61316681e+00 -8.67320824e+00 -8.12954044e+00]]

  [[ 1.18172336e+00  8.86805296e-01  1.63558555e+00  4.20765400e+00
     4.41581488e+00 -4.85230350e+00 -7.42733288e+00]
   [ 4.40815359e-01 -4.11654329e+00 -2.52238703e+00  3.83199716e+00
     3.61510658e+00 -5.21432638e+00  6.40418005e+00]
   [-1.39663115e+01  5.36948872e+00  1.69813246e-01  5.84380329e-01
    -2.94225430e+00  1.31688917e+00 -1.10627632e+01]]]]
#输出
******************** op8********************

-------------------------------------------------------------------------------------------------  
init = tf.global_variables_initializer() 
with tf.Session() as sess:
    sess.run(init)
    print('*' * 20 + ' op1 ' + '*' * 20)
    print(sess.run(op1))
    print('*' * 20 + ' op2 ' + '*' * 20)
    print(sess.run(op2))
    print('*' * 20 + ' op3 ' + '*' * 20)
    print(sess.run(op3))
    print('*' * 20 + ' op4 ' + '*' * 20)
    print(sess.run(op4))
    print('*' * 20 + ' op5 ' + '*' * 20)
    print(sess.run(op5))
    print('*' * 20 + ' op6 ' + '*' * 20)
    print(sess.run(op6))
    print('*' * 20 + ' op7 ' + '*' * 20)
    print(sess.run(op7))
    print('*' * 20 + ' op8 ' + '*' * 20)
    print(sess.run(op8))
-------------------------------------------------------------------------------------------------
原文：https://blog.csdn.net/zuolixiangfisher/article/details/80528989
Tensorflow——理解tf.nn.conv2d方法

定义

举例

相关文章

网友评论

延伸阅读

深度阅读

栏目导航

热点阅读

python与Tensorflow
