tf卷積基礎

slim模組的conv2d函式，是二維卷積介面，順著源**可以看到最終調的tensorflow介面是convolution，這個地方就進入c++層面了，暫時不涉及。先來看看這個convolution函式，官方定義是這樣的：

tf.nn.convolution(
input,
filter,
padding,
strides=none,
dilation_rate=none,
name=none,
data_format=none
)

一般情況下，也就是data_format=none的時候，input的要求格式是[batch_size] + input_spatial_shape + [in_channels], 也就是要求第一維是batch，最後一維是channel，中間是真正的卷積維度。所以這個介面不僅只支援2維卷積，猜測2維卷積tf.nn.conv2d是對此介面的封裝。[batch, height, weight, channel]就是conv2d的input引數格式，batch就是樣本數，或者更狹隘一點，數量，height是高，weight是的寬，slim的分類網路都是height=weight的，以實現方陣運算，所有slim模組中的原始都需要經過預處理過程，這裡不展開。

filter引數是卷積核的定義，spatial_filter_shape + [in_channels, out_channels]，對於2維卷積同樣是4維引數[weight, height, channel, out_channel]

例子1：單通道單卷積核

import tensorflow as tf

input = tf.constant(

[ [

[[100.],

[70.],

[40.]],[

[90.],

[60.],

[30.]],[

[80.],

[50.],

[20.]

]]])

#定義乙個2x2x1x1的卷積核

filter = tf.constant(

[ [

[[0.5]],[

[0.2]]],

[[[0.1]],[

[0.3]

]]])

#即卷積核為： [0.5,0.1]

# [0.2,0.3]

result = tf.nn.convolution(input, filter, padding='valid');

with tf.session() as sess:

print sess.run(result)

即輸入為：

[100,90,80]

[70,60,50]

[40,30,20]

卷積結果

[91,80]
[58,47]

例子2：三通道單卷積核

import tensorflow as tf

input = tf.constant(

[ [

[[100., 100., 20.],

[100., 100., 100.],

[100., 100., 100.]],[

[60., 100., 100.],

[100., 100., 100.],

[100., 100., 100.]],[

[20., 100., 100.],

[100., 100., 100.],

]]])

#定義2x2x3x1的卷積核（3為通道數，1為輸出特徵圖數）

filter = tf.constant(

[ [

[[0.5],

[0.5],

[0.5]],[

[0.1],

[0.5],

[0.5]]],

[[[0.2],

[0.5],

[0.5]],[

[0.3],

[0.5],

[0.5]

]],]);

result = tf.nn.convolution(input, filter, padding='valid');

with tf.session() as sess:

print sess.run(result)

」』即三通道為：

[100,60,20] 。。 [100,100,100] 。 [20,100,100]

[100,100,100] 。 [100,100,100] 。 [100,100,100]

」』對應通道的卷積核為：

[0.5,0.2] ….[0.5,0.5] ….[0.5,0.5]

[0.1,0.3] ….[0.5,0.5] ….[0.5,0.5]

結果：

[462.474]
[510.510]

例子3:三通道2個卷積核

#定義乙個1x3x3x3的
input = tf.constant(
[ [
[[100., 100., 20.],
[100., 100., 100.],
[100., 100., 100.]],[
[60., 100., 100.],
[100., 100., 100.],
[100., 100., 100.]],[
[20., 100., 100.],
[100., 100., 100.],
[100., 100., 100.],
]]])
#定義2x2x3x2的卷積核
filter2 = tf.constant(
[ [
[[0.5, 0.2],
[0.5, 0.1],
[0.5, 0.1]],[
[0.4, 0.4],
[0.5, 0.2],
[0.5, 0.1]]],
[[[0.3, 0.3],
[0.5, 0.1],
[0.5, 0.1]],[
[0.2, 0.1],
[0.5, 0.1],
[0.5, 0.1]]],
])

卷積核為：

[0.5,0.2] ….[0.5,0.5] ….[0.5,0.5]

[0.1,0.3] ….[0.5,0.5] ….[0.5,0.5]

[0.2,0.3] ….[0.1,0.1] ….[0.10.1]

[0.4,0.1 ….[0.2,0.1] ….[0.1,0.1]

結果：

[462,474]
[510,510]
[170,158]
[190,190]

tf卷積基礎

tf基礎操作

tf基礎知識

tf 矩陣行和列交換 tf矩陣基礎

tf卷積基礎

tf基礎操作

tf基礎知識

tf 矩陣行和列交換 tf矩陣基礎

相關推薦