caffe中新增自己的激活函数层_C++实现

一、前言

本篇主要转载自一个视频教程，主要实现在caffe中新增自己的层。

二、具体做法

自定义一个计算层，实现y=x^power+bur的功能，事实上这个新层为激活函数层

三、实现的方法思路

(1)任何一个层都可以被继承，然后进行重写函数

(2)尽量确保要实现的功能是否必须要自己写，不然尽量用已有的层，每一个层在caffe/include/caffe/layers
源码中都有详细的介绍

四、整体步骤

1. 创建新定义的头文件include/caffe/layers/my_neuron_layer.hpp
重新Layer名的方法：virtual inline const char* type() const { return "MyNeuron"; }
如果只是需要cpu方法的话，可以注释掉forward/backward_gpu()这两个方法
2. 创建对应src/caffe/src/my_neuron_layer.cpp的源文件
重写方法LayerSetUp,实现从能从prototxt读取参数，这个没有从prototxt读取新的参数，则不需要重写
重写方法Reshape，如果对继承类没有修改的话，就不需要重写
重写方法Forward_cpu
重写方法Backward_cpu(非必须)
如果要GPU支持，则还需要创建src/caffe/src/my_neuron_layer.cu，同理重写方法Forward_gpu/Backward_gpu(非必须)
3. 到proto/caffe.proto 注册新的layer
4. 在my_neuron_layer.cpp添加注册的宏定义
INSTANTIATE_CLASS(MyNeuronLayer);
REGISTER_LAYER_CLASS(MyNeuron);
如果有my_neuron_layer.cu 文件，则还要添加
INSTANTIATE_LAYER_GPU_FUNCS(MyNeuronLayer);
5. 重新make和install

五、具体做法（仅仅用CPU）

5.1、prototxt文件中新层的描述

由于需要从prototxt文件中读取参数初始化 power 和 bur，所实现新层为激活函数的功能，故可参考sigmoid_layer 的写法，先参看sigmoid在prototxt文件的描述：

layer {
name: "sigmoid"
type: "Sigmoid"
bottom: "fc1"
top: "fc1"
}

新增参数 power 和 bur，在prototxt文件中描述可这样写：

layer {
  name: "myneuron"
  type: "MyNeuron"  #name和type不一样
  bottom: "fc1"
  top: "fc1"
  my_neuron_param {
    power: 3
    bur: 1
  }
}

5.2、新增 myNeuron_layer.hpp 和 myNeuron_layer.cpp

方便的做法是参考相似层的写法，这里将sigmoid_layer.hpp和sigmoid_layer.cpp 改为相应名字后，分别保存至include/caffe/layers/ 和 src/caffe/layers/ , 并分别修改如下：

//myNeuron_layer.hpp
#ifndef CAFFE_MY_NEURON_LAYER_HPP_
#define CAFFE_MY_NEURON_LAYER_HPP_

#include <vector>

#include "caffe/blob.hpp"
#include "caffe/layer.hpp"
#include "caffe/proto/caffe.pb.h"

#include "caffe/layers/neuron_layer.hpp"

namespace caffe {

template <typename Dtype>
class MyNeuronLayer : public NeuronLayer<Dtype> {
 public:
  
  explicit MyNeuronLayer(const LayerParameter& param)
      : NeuronLayer<Dtype>(param) {}
  virtual void LayerSetUp(const vector<Blob<Dtype>*>& bottom,
      const vector<Blob<Dtype>*>& top);

  virtual inline const char* type() const { return "MyNeuron"; }

 protected:
  
  virtual void Forward_cpu(const vector<Blob<Dtype>*>& bottom,
      const vector<Blob<Dtype>*>& top);
  virtual void Forward_gpu(const vector<Blob<Dtype>*>& bottom,const vector<Blob<Dtype>*>& top);

  virtual void Backward_cpu(const vector<Blob<Dtype>*>& top,
      const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom);
  virtual void Backward_gpu(const vector<Blob<Dtype>*>& top,const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom);

  Dtype power_;
  Dtype bur_;
};

}  // namespace caffe

#endif  // CAFFE_MY_NEURON_LAYER_HPP_

//myNeuron_layer.cpp
#include <vector>

#include "caffe/layers/myNeuron_layer.hpp"
#include "caffe/util/math_functions.hpp"

namespace caffe {

template <typename Dtype>
void MyNeuronLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,const vector<Blob<Dtype>*>& top){
  
  NeuronLayer<Dtype>::LayerSetUp(bottom,top);
  power_ = this->layer_param_.my_neuron_param().power();
  bur_ = this->layer_param_.my_neuron_param().bur();
}

// Compute y = x^power
template <typename Dtype>
void MyNeuronLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,const vector<Blob<Dtype>*>& top){

  Dtype* top_data = top[0]->mutable_cpu_data();
  const int count = bottom[0]->count();
  caffe_powx(count, bottom[0]->cpu_data(), Dtype(power_), top_data);
  caffe_add_scale(count, Dtype(bur_), top_data);
}

template <typename Dtype>
void MyNeuronLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,const vector<bool>& propagate_down,const vector<Blob<Dtype>*>& bottom){
  const int count = top[0]->count();
  const Dtype* top_diff = top[0]->cpu_diff();
  if(propagate_down[0]){
    const Dtype* bottom_data = bottom[0]->cpu_data();
    Dtype* bottom_diff = bottom[0]->mutable_cpu_diff();
    caffe_powx(count, bottom_data, Dtype(power_ - 1), bottom_diff);
    caffe_scal(count, Dtype(power_), bottom_diff);
    caffe_mul(count, bottom_diff, top_diff, bottom_diff);
  }

}

#ifdef CPU_ONLY
STUB_GPU(MyNeuronLayer);
#endif

INSTANTIATE_CLASS(MyNeuronLayer);  
REGISTER_LAYER_CLASS(MyNeuron);

}// namespace caffe

5.3、注册
接下来需要进行注册。

(1) 到src/caffe/proto/caffe.proto 注册 prototxt中的 my_neuron_param

message LayerParameter {
...
optional MyNeuronParameter my_neuron_param = 150;
...
}
...
message MyNeuronParameter{
  optional float power = 1 [default = 2];
  optional float bur = 2 [default = 1];
}
...
message V1LayerParameter {
...
MYNEURON = 40;  #大写
...
}

V1layer注册的是全大写的类型名字，这点有点不知为何
(2) 在cpp文件中最后注册cpp中所创建的新类和相应 prototxt中的类型名

INSTANTIATE_CLASS(MyNeuronLayer);  
REGISTER_LAYER_CLASS(MyNeuron);

5.4、测试

重新make之后，创建deploy.prototxt 和 test_my_neuron.py 如下

name: "CaffeNet"
input: "data"
input_shape {
  dim: 1 # batchsize
  dim: 1 # number of colour channels - rgb
  dim: 28 # width
  dim: 28 # height
}
layer {
  name: "myneuron"
  type: "MyNeuron"
  bottom: "data"
  top: "data_out"
  my_neuron_param {
    power : 2
    bur : 1
  }
}

# -*- coding: utf-8 -*-

import numpy as np
import matplotlib.pyplot as plt
import os
import sys

deploy_file = "./deploy.prototxt"
test_data   = "./5.jpg"

if __name__ == '__main__':
  sys.path.append("/home/zjy/caffe/python")
  import caffe

  net = caffe.Net(deploy_file,caffe.TEST)

  transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})

  transformer.set_transpose('data', (2, 0, 1))

  img = caffe.io.load_image(test_data,color=False)

  net.blobs['data'].data[...] = transformer.preprocess('data', img)

  print net.blobs['data'].data[0][0][14]

  out = net.forward()

  print out['data_out'][0][0][14]

5.5、运行结果

六、具体做法（用GPU）

6.1、prototxt文件中新层的描述

保持不变

6.2、新增 myNeuron_layer.hpp 和 myNeuron_layer.cu

myNeuron_layer.hpp 保持不变，在src/caffe/layers/ 增加 myNeuron_layer.cu

#include <vector>

#include "caffe/layers/myNeuron_layer.hpp"
#include "caffe/util/math_functions.hpp"
#include <iostream>
using namespace std;

namespace caffe {

template <typename Dtype>
void MyNeuronLayer<Dtype>::Forward_gpu(
    const vector<Blob<Dtype>*>& bottom, const vector<Blob<Dtype>*>& top) {
  const int count = top[0]->count();
  Dtype* top_data = top[0]->mutable_gpu_data();
  caffe_gpu_powx(count, bottom[0]->gpu_data(), Dtype(power_), top_data);
}

template <typename Dtype>
void MyNeuronLayer<Dtype>::Backward_gpu(const vector<Blob<Dtype>*>& top,
    const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) {
  const int count = top[0]->count();
  const Dtype* top_diff = top[0]->gpu_diff();
  if (propagate_down[0]) {
    const Dtype* bottom_data = bottom[0]->gpu_data();
    Dtype* bottom_diff = bottom[0]->mutable_gpu_diff();
    const Dtype* bottom_data_w = bottom[0]->cpu_data();
    const Dtype* bottom_diff_w = bottom[0]->cpu_diff();

    cout << "bottom_data[0]: " << bottom_data_w[0] << endl;
    cout << "bottom_diff[0]: " << bottom_diff_w[0] << endl;

    caffe_gpu_powx(count, bottom_data, Dtype(power_ - 1), bottom_diff);

    bottom_diff = bottom[0]->mutable_gpu_diff();
    bottom_data_w = bottom[0]->cpu_data();
    bottom_diff_w = bottom[0]->cpu_diff();
    cout << "bottom_data[0]: " << bottom_data_w[0] << endl;
    cout << "bottom_diff[0]: " << bottom_diff_w[0] << endl;

    caffe_gpu_scal(count, Dtype(power_), bottom_diff);

    bottom_diff = bottom[0]->mutable_gpu_diff();
    bottom_data_w = bottom[0]->cpu_data();
    bottom_diff_w = bottom[0]->cpu_diff();
    cout << "bottom_data[0]: " << bottom_data_w[0] << endl;
    cout << "bottom_diff[0]: " << bottom_diff_w[0] << endl;

    caffe_gpu_mul(count, bottom_diff, top_diff, bottom_diff);

    bottom_diff = bottom[0]->mutable_gpu_diff();
    bottom_data_w = bottom[0]->cpu_data();
    bottom_diff_w = bottom[0]->cpu_diff();
    cout << "bottom_data[0]: " << bottom_data_w[0] << endl;
    cout << "bottom_diff[0]: " << bottom_diff_w[0] << endl;
  }
}

INSTANTIATE_LAYER_GPU_FUNCS(MyNeuronLayer);

}  // namespace caffe

5.3、注册
(1) 保持不变

(2) 在 cu 文件后添加

INSTANTIATE_LAYER_GPU_FUNCS(MyNeuronLayer);

5.4、测试
重新 make之后，创建 deploy.prototxt 和 test_my_neuron_gpu.py , deploy.prototxt 保持不变，test_my_neuron_gpu.py

如下：

# -*- coding: utf-8 -*-

import numpy as np
import matplotlib.pyplot as plt
import os
import sys

deploy_file = "./deploy.prototxt"
test_data   = "./5.jpg"

if __name__ == '__main__':
  sys.path.append("/home/zjy/caffe/python")
  import caffe
  caffe.set_mode_gpu()

  net = caffe.Net(deploy_file,caffe.TEST)

  transformer = caffe.io.Transformer({'data': net.blobs['data'].data.shape})

  transformer.set_transpose('data', (2, 0, 1))

  img = caffe.io.load_image(test_data,color=False)

  net.blobs['data'].data[...] = transformer.preprocess('data', img)

  print net.blobs['data'].data[0][0][14]

  out = net.forward()

  print out['data_out'][0][0][14]

caffe中新增自己的激活函数层_C++实现

猜你喜欢