NCHW转化为NHWC

NCHW转化为NHWC：

NHWC：[batch,height,width,channels]

NCHW：[batch,channels,height,width]

Tensorflow中使用的格式为NHWC：

API:
tf.transpose(a,perm)

• a:需要转置的张量。
• perm:按照哪种排列组合进行转置。

实例：

import tensorflow as tf

#定义一个常量 是一个三阶张量，[channels,height,width] [0,1,2]
nchw = tf.constant(value=[1,2,3,4,5,6,7,8,9,10,11,12],shape=[3,2,2])

#将[channels,height,width]转置为[height,width，channels]格式，[0,1,2] ->[1,2,0]
nhwc = tf.transpose(nchw,[1,2,0])


with tf.Session() as sess:

    nchw_new,nhwc_new = sess.run([nchw,nhwc])
    print("nchw:\n",nchw_new)
    print("nhwc:\n",nhwc_new)

'''
nchw:
 [[[ 1  2]
  [ 3  4]]

 [[ 5  6]
  [ 7  8]]

 [[ 9 10]
  [11 12]]]
nhwc:
 [[[ 1  5  9]
  [ 2  6 10]]

 [[ 3  7 11]
  [ 4  8 12]]]

'''