目录

　　举例

　　单个张量与多个卷积核的分离卷积

　　参考资料

分离卷积就是先在深度上分别卷积，然后再进行卷积，对应代码为：

import tensorflow as tf

# [batch, in_height, in_width, in_channels]
input =tf.reshape(tf.constant([2,5,3,3,8,2,6,1,1,2,5,4,7,9,2,3,-1,3], tf.float32),[1,3,3,2])

# [filter_height, filter_width, in_channels, out_channels]
depthwise_filter = tf.reshape(tf.constant([3,1,-2,2,-1,-3,4,5], tf.float32),[2,2,2,1])
pointwise_filter = tf.reshape(tf.constant([-1,1], tf.float32),[1,1,2,1])

print(tf.Session().run(tf.nn.separable_conv2d(input,depthwise_filter,pointwise_filter,[1,1,1,1],"VALID")))
[[[[ 20.]
   [  9.]]

  [[-24.]
   [ 29.]]]]

返回目录

对应代码为：

import tensorflow as tf

# [batch, in_height, in_width, in_channels]
input =tf.reshape(tf.constant([2,5,3,3,8,2,6,1,1,2,5,4,7,9,2,3,-1,3], tf.float32),[1,3,3,2])

# [filter_height, filter_width, in_channels, out_channels]
depthwise_filter = tf.reshape(tf.constant([3,1,-3,1,-1,7,-2,2,-5,2,7,3,-1,3,1,-3,-8,6,4,6,8,5,9,-5], tf.float32),[2,2,2,3])
pointwise_filter = tf.reshape(tf.constant([0,0,1,0,0,1,0,0,0,0,0,0], tf.float32),[1,1,6,2])

print(tf.Session().run(tf.nn.separable_conv2d(input,depthwise_filter,pointwise_filter,[1,1,1,1],"VALID")))
[[[[ 32.  -7.]
   [ 52.  -8.]]

  [[ 41.   0.]
   [ 11. -34.]]]]

返回目录

《图解深度学习与神经网络：从张量到TensorFlow实现》_张平

返回目录

原文地址：https://www.cnblogs.com/itmorn/p/11250848.html