一.Concate
作用:实现多个输入数据的拼接
输入:x1,x2,…,xk 输出:y
x1: N*C*H*W
x2: N*C*H*W
xk: N*C*H*W
y: kN*C*H*W(concate_dim=0)
y: N*kC*H*W(concate_dim=1)
参数:两个作用相同 ①axis ②concate_dim
(1)caffe.proto层参数定义
message ConcatParameter {
//指定拼接的维度,默认为1:channel;支持负索引:-1表示最后一个维度。
optional int32 axis = 2 [default = 1];
//作用同axis一样,但不能指定为负索引。
optional uint32 concat_dim = 1 [default = 1];
}
(2)LayerSetUp
template <typename Dtype>
void ConcatLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
//给出axis以及concate_dim参数信息
const ConcatParameter& concat_param = this->layer_param_.concat_param();
CHECK(!(concat_param.has_axis() && concat_param.has_concat_dim()))
<< "Either axis or concat_dim should be specified; not both.";
}
(3)Reshape
template <typename Dtype>
void ConcatLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
//获取输入维度:num_axis
//获取拼接参数:concate_param
const int num_axes = bottom[0]->num_axes();
const ConcatParameter& concat_param = this->layer_param_.concat_param();
if (concat_param.has_concat_dim()) {
//获取指定拼接维度:concate_dim
concat_axis_ = static_cast<int>(concat_param.concat_dim());
// Don't allow negative indexing for concat_dim, a uint32 -- almost
// certainly unintended.
//CHECK_GE(≥):拼接维度≥0,不支持负索引。
CHECK_GE(concat_axis_, 0) << "casting concat_dim from uint32 to int32 "
<< "produced negative result; concat_dim must satisfy "
<< "0 <= concat_dim < " << kMaxBlobAxes;
//CHECK_LT(<):拼接维度<输入维度
CHECK_LT(concat_axis_, num_axes) << "concat_dim out of range.";
} else {否则,以channel维度进行拼接
//标准化参数索引
concat_axis_ = bottom[0]->CanonicalAxisIndex(concat_param.axis());
}
// Initialize with the first blob.
//初始化输出top_shape与输入shape大小一致
vector<int> top_shape = bottom[0]->shape();//输出top和输入bottom的shape相同
num_concats_ = bottom[0]->count(0, concat_axis_);//拼接的数目:num_concats_
concat_input_size_ = bottom[0]->count(concat_axis_ + 1);
int bottom_count_sum = bottom[0]->count();
for (int i = 1; i < bottom.size(); ++i) {
CHECK_EQ(num_axes, bottom[i]->num_axes())//判断输入维度num_axes是否一致
<< "All inputs must have the same #axes.";
for (int j = 0; j < num_axes; ++j) {
//除去待拼接的维度concate_axis_外,其他维度大小是否相同
if (j == concat_axis_) { continue; }
CHECK_EQ(top_shape[j], bottom[i]->shape(j))
<< "All inputs must have the same shape, except at concat_axis.";
}
bottom_count_sum += bottom[i]->count();//累加第i个输入的个数
top_shape[concat_axis_] += bottom[i]->shape(concat_axis_);//累加到输出top
}
top[0]->Reshape(top_shape);//Reshape输出top
CHECK_EQ(bottom_count_sum, top[0]->count());
if (bottom.size() == 1) {
//只有一个输入bottom[0],直接复制成输出top[0]
//梯度shape也和输入bottom[0]一致
top[0]->ShareData(*bottom[0]);
top[0]->ShareDiff(*bottom[0]);
}
}
(2)Forward前向传播
template <typename Dtype>
void ConcatLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
if (bottom.size() == 1) { return; }
Dtype* top_data = top[0]->mutable_cpu_data();
int offset_concat_axis = 0;
const int top_concat_axis = top[0]->shape(concat_axis_);
for (int i = 0; i < bottom.size(); ++i) {//遍历所有bottom
const Dtype* bottom_data = bottom[i]->cpu_data();
const int bottom_concat_axis = bottom[i]->shape(concat_axis_);
for (int n = 0; n < num_concats_; ++n) {
//bottom复制给top
caffe_copy(bottom_concat_axis * concat_input_size_,
bottom_data + n * bottom_concat_axis * concat_input_size_,
top_data + (n * top_concat_axis + offset_concat_axis)
* concat_input_size_);
}
offset_concat_axis += bottom_concat_axis;
}
}
(3) Backward反向传播
原理:把输出求得的梯度直接复制给输入即可。
template <typename Dtype>
void ConcatLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) {
if (bottom.size() == 1) { return; }
const Dtype* top_diff = top[0]->cpu_diff();//梯度
int offset_concat_axis = 0;
const int top_concat_axis = top[0]->shape(concat_axis_);
for (int i = 0; i < bottom.size(); ++i) {
const int bottom_concat_axis = bottom[i]->shape(concat_axis_);
if (propagate_down[i]) {
Dtype* bottom_diff = bottom[i]->mutable_cpu_diff();
for (int n = 0; n < num_concats_; ++n) {
caffe_copy(bottom_concat_axis * concat_input_size_, top_diff +
(n * top_concat_axis + offset_concat_axis) * concat_input_size_,
bottom_diff + n * bottom_concat_axis * concat_input_size_);
}
}
offset_concat_axis += bottom_concat_axis;
}
}
(4)Usage使用
layer{
name: "feat_all"
type: "Concat"
bottom: "feat"//输入a
bottom: "feat_p"//输入b
top: "feat_all"//concate输出
concat_param {
axis: 1//在channel维度进行拼接
}
}
二.Slice
作用:把输入按维度进行切片
(1)caffe.proto层参数定义
参数:①axis ②slice_point ③slice_dim
message SliceParameter {
optional int32 axis = 3 [default = 1];//要进行切片的维度
repeated uint32 slice_point = 2;//将维度axis按照slice_point进行切分
optional uint32 slice_dim = 1 [default = 1];//作用同axis
}
同样包括LayerSetUp、Reshape、Forward、Backard
(1)LayerSetUp
template <typename Dtype>
void SliceLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
//获取slice参数(slice_param):axis和slice_dim
const SliceParameter& slice_param = this->layer_param_.slice_param();
CHECK(!(slice_param.has_axis() && slice_param.has_slice_dim()))
<< "Either axis or slice_dim should be specified; not both.";
slice_point_.clear();
std::copy(slice_param.slice_point().begin(),
slice_param.slice_point().end(),
std::back_inserter(slice_point_));
}
(2)Reshape
template <typename Dtype>
void SliceLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
const int num_axes = bottom[0]->num_axes();//输入维度数目
const SliceParameter& slice_param = this->layer_param_.slice_param();
if (slice_param.has_slice_dim()) {
//若指定切片维度,则在相应维度上进行切片
slice_axis_ = static_cast<int>(slice_param.slice_dim());
// Don't allow negative indexing for slice_dim, a uint32 -- almost
// certainly unintended.
//0≤切片维度<输入维度数目
CHECK_GE(slice_axis_, 0) << "casting slice_dim from uint32 to int32 "
<< "produced negative result; slice_dim must satisfy "
<< "0 <= slice_dim < " << kMaxBlobAxes;
CHECK_LT(slice_axis_, num_axes) << "slice_dim out of range.";
//若没有指定切片维度,则在channel维度上进行切分
//标准化参数索引
slice_axis_ = bottom[0]->CanonicalAxisIndex(slice_param.axis());
}
vector<int> top_shape = bottom[0]->shape();
//获取切片维度slice_axis_大小shape(.)
const int bottom_slice_axis = bottom[0]->shape(slice_axis_);
//计算每次切分在一个batch中需要切分的次数,若在channel维进行切分,等于batchsize
num_slices_ = bottom[0]->count(0, slice_axis_);
//切分后feature map的大小,若在channel维度切分,等于H×W
slice_size_ = bottom[0]->count(slice_axis_ + 1);
int count = 0;
if (slice_point_.size() != 0) {
CHECK_EQ(slice_point_.size(), top.size() - 1);
CHECK_LE(top.size(), bottom_slice_axis);
int prev = 0;
vector<int> slices;
for (int i = 0; i < slice_point_.size(); ++i) {
CHECK_GT(slice_point_[i], prev);
//统计每次切分的跨度
slices.push_back(slice_point_[i] - prev);
prev = slice_point_[i];
}
slices.push_back(bottom_slice_axis - prev);
//根据切分跨度,对输出top_shape进行Reshape,并统计每个输出切分维度的大小
for (int i = 0; i < top.size(); ++i) {
top_shape[slice_axis_] = slices[i];
top[i]->Reshape(top_shape);
count += top[i]->count();
}
} else {
CHECK_EQ(bottom_slice_axis % top.size(), 0)
<< "Number of top blobs (" << top.size() << ") should evenly "
<< "divide input slice axis (" << bottom_slice_axis << ")";
top_shape[slice_axis_] = bottom_slice_axis / top.size();
for (int i = 0; i < top.size(); ++i) {
top[i]->Reshape(top_shape);
count += top[i]->count();
}
}
CHECK_EQ(count, bottom[0]->count());
if (top.size() == 1) {
top[0]->ShareData(*bottom[0]);
top[0]->ShareDiff(*bottom[0]);
}
}
(3)Forward前向传播
template <typename Dtype>
void SliceLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
//只有一个输出数据,不进行切分
if (top.size() == 1) { return; }
int offset_slice_axis = 0;
//获取输入数据bottom_data
const Dtype* bottom_data = bottom[0]->cpu_data();
//获取要切分的维度slice_axis_大小shape(.)
const int bottom_slice_axis = bottom[0]->shape(slice_axis_);
for (int i = 0; i < top.size(); ++i) {
Dtype* top_data = top[i]->mutable_cpu_data();//对输出top_data进行slice
const int top_slice_axis = top[i]->shape(slice_axis_);//要切分的维度大小
for (int n = 0; n < num_slices_; ++n) {
const int top_offset = n * top_slice_axis * slice_size_;
const int bottom_offset =
(n * bottom_slice_axis + offset_slice_axis) * slice_size_;
//拷贝数据
caffe_copy(top_slice_axis * slice_size_,
bottom_data + bottom_offset, top_data + top_offset);
}
//移动下一个切分点
offset_slice_axis += top_slice_axis;
}
}
(4)Backward反向传播
template <typename Dtype>
void SliceLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) {
if (!propagate_down[0] || top.size() == 1) { return; }
int offset_slice_axis = 0;
Dtype* bottom_diff = bottom[0]->mutable_cpu_diff();
const int bottom_slice_axis = bottom[0]->shape(slice_axis_);
for (int i = 0; i < top.size(); ++i) {
const Dtype* top_diff = top[i]->cpu_diff();
const int top_slice_axis = top[i]->shape(slice_axis_);
for (int n = 0; n < num_slices_; ++n) {
//将top diff按照切分大小分别拷贝到bottom_diff中,每次移动top_slice_axis = 4个feature map
const int top_offset = n * top_slice_axis * slice_size_;
//而bottom diff则每次要移动16个feature map,对应不同的输入
const int bottom_offset =
(n * bottom_slice_axis + offset_slice_axis) * slice_size_;
//拷贝数据
caffe_copy(top_slice_axis * slice_size_,
top_diff + top_offset, bottom_diff + bottom_offset);
}
//移动下一个切分点
//bottom diff指针移动到下一个切分点,然后继续从top diff中拷贝
offset_slice_axis += top_slice_axis;
}
}
(5)Usage使用
说明:slice_point的个数等于top个数减去1
layer {
name: "slice_pair"
type: "Slice"
bottom: "pair_data"
top: "data"
top: "data_p"
slice_param {
slice_dim: 1 //在channel维度上进行slice
slice_point: 1
}
}