tensorflow.squeeze(input, squeeze_dims=None, name=None)
参数: input --> 输入的tensor
squeeze_dims = None -->默认None是删除input中所有大小是1的维度,若指定位置则删除所指定位置大小是1的维度
name -->名称(可选)
原始数据
y = tf.expand_dims(y,axis=-1)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
value = sess.run(y)
print (value)
print (y.shape)
[[[[[[1]
[2]
[3]]]
[[[4]
[5]
[6]]]]]]
(1, 1, 2, 1, 3, 1)
删除所有大小是1的维度:
z = tf.squeeze(y)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
value = sess.run(z)
print (value)
print (z.shape)
print ("z[0][1]: ",value[0][1])
[[1 2 3]
[4 5 6]]
(2, 3)
z[0][1]: 2
删除位置是3,5的大小是1的维度(从0起)
z1 = tf.squeeze(y, [3, 5])
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
value = sess.run(z1)
print (value)
print (z1.shape)
[[[[1 2 3]
[4 5 6]]]]
(1, 1, 2, 3)