提交 b3db8ef8 编写于 作者: P peterzhang2029

refine the description

上级 a4742115
......@@ -2545,11 +2545,17 @@ def img_conv_layer(input,
num_filters.
There are several groups of filters in PaddlePaddle implementation.
Each group will process some channels of the input. For example, if
num_channel = 256, group = 4, num_filter=32, the PaddlePaddle will create
32 filters to process the input. The input channels will be split into 4
pieces. First 256/4 = 64 channels will be processed by first 32/4 = 8 filters.
The rest channels will be processed by the rest groups of filters.
If the groups attribute is greater than 1, for example groups=2,
the input will be splitted into 2 parts along the channel axis, and
the filters will also be splitted into 2 parts. The first half of the filters
is only connected to the first half of the input channels, while the second
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.
The details of grouped convolution, please refer to:
`ImageNet Classification with Deep Convolutional Neural Networks
<http://www.cs.toronto.edu/~kriz/imagenet_classification_with_deep_convolutional.pdf>`_
The example usage is:
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册