@@ -518,13 +518,17 @@ please refer to the following explanation and references therein
...
@@ -518,13 +518,17 @@ please refer to the following explanation and references therein
what-are-deconvolutional-layers/>`_ .
what-are-deconvolutional-layers/>`_ .
The num_channel means input image’s channel number. It may be 1 or 3 when
The num_channel means input image’s channel number. It may be 1 or 3 when
input is raw pixels of image(mono or RGB), or it may be the previous layer’s
input is raw pixels of image(mono or RGB), or it may be the previous layer’s
num_filters * num_group.</p>
num_filters.</p>
<p>There are several groups of filters in PaddlePaddle implementation.
<p>There are several groups of filters in PaddlePaddle implementation.
Each group will process some channels of the input. For example, if
If the groups attribute is greater than 1, for example groups=2,
num_channel = 256, group = 4, num_filter=32, the PaddlePaddle will create
the input will be splitted into 2 parts along the channel axis, and
32*4 = 128 filters to process the input. The channels will be split into 4
the filters will also be splitted into 2 parts. The first half of the filters
pieces. First 256/4 = 64 channels will be processed by first 32 filters. The
is only connected to the first half of the input channels, while the second
rest channels will be processed by the rest groups of filters.</p>
half of the filters is only connected to the second half of the input. After
the computation of convolution for each part of input,
the output will be obtained by concatenating the two results.</p>
<p>The details of grouped convolution, please refer to:
<aclass="reference external"href="http://www.cs.toronto.edu/~kriz/imagenet_classification_with_deep_convolutional.pdf">ImageNet Classification with Deep Convolutional Neural Networks</a></p>
@@ -531,13 +531,17 @@ please refer to the following explanation and references therein
...
@@ -531,13 +531,17 @@ please refer to the following explanation and references therein
what-are-deconvolutional-layers/>`_ .
what-are-deconvolutional-layers/>`_ .
The num_channel means input image’s channel number. It may be 1 or 3 when
The num_channel means input image’s channel number. It may be 1 or 3 when
input is raw pixels of image(mono or RGB), or it may be the previous layer’s
input is raw pixels of image(mono or RGB), or it may be the previous layer’s
num_filters * num_group.</p>
num_filters.</p>
<p>There are several groups of filters in PaddlePaddle implementation.
<p>There are several groups of filters in PaddlePaddle implementation.
Each group will process some channels of the input. For example, if
If the groups attribute is greater than 1, for example groups=2,
num_channel = 256, group = 4, num_filter=32, the PaddlePaddle will create
the input will be splitted into 2 parts along the channel axis, and
32*4 = 128 filters to process the input. The channels will be split into 4
the filters will also be splitted into 2 parts. The first half of the filters
pieces. First 256/4 = 64 channels will be processed by first 32 filters. The
is only connected to the first half of the input channels, while the second
rest channels will be processed by the rest groups of filters.</p>
half of the filters is only connected to the second half of the input. After
the computation of convolution for each part of input,
the output will be obtained by concatenating the two results.</p>
<p>The details of grouped convolution, please refer to:
<aclass="reference external"href="http://www.cs.toronto.edu/~kriz/imagenet_classification_with_deep_convolutional.pdf">ImageNet Classification with Deep Convolutional Neural Networks</a></p>
