ShuffleNet-master
所属分类：人工智能/神经网络/深度学习
开发工具：C++
文件大小：6KB
下载次数：1
上传日期：2017-09-23 19:43:24
上 传 者：
wu joy
说明： 一种专门为移动端设备而设计的高效卷积神经网络结构——ShuffleNet，这种新结构将点态组卷积（pointwise group convolution）和通道随机混合（channel shuffle）这两种经典方法进行结合与改进，大大提升了计算效率 。
(ShuffleNet is an efficient convolutional neural network designed for mobile terminal devices. This new structure combines and improves the two classical methods, pointwise group convolution and channel shuffle, which greatly improves the computational efficiency.)
文件列表:
shuffle_channel_layer.cpp (3293, 2017-08-02)
shuffle_channel_layer.cu (3425, 2017-08-02)
shuffle_channel_layer.hpp (1595, 2017-08-02)
shufflenet_1x_g3_deploy.prototxt (47710, 2017-08-02)
# ShuffleNet
This is caffe implementation of ShuffleNet, For details, please read the original paper:
["ShuffleNet: An Extremely Efficient Convolutional
Neural Network for Mobile Devices" by Xiangyu Zhang et. al. 2017](https://arxiv.org/pdf/1707.01083.pdf).
This code is based on camel007's implementation(https://github.com/camel007/Caffe-ShuffleNet), but I recode the cuda file for acceleration.
## How to use?
#### caffe.proto:
```
message LayerParameter {
...
optional ShuffleChannelParameter shuffle_channel_param = 1***;
...
}
...
message ShuffleChannelParameter {
optional uint32 group = 1[default = 1]; // The number of group
}
```
近期下载者：
相关文件：
收藏者：