1. RoiPooling层是对Caffe层的扩展,目的:把ROI区域归一化到相同尺寸大小,以便于后面的全连接层处理。
2. 部分网络协议:
layer {
name: "conv5"
type: "Convolution"
bottom: "conv4"
top: "conv5"
param { lr_mult: 1.0 }
param { lr_mult: 2.0 }
convolution_param {
num_output: 256
kernel_size: 3
pad: 1
stride: 1
}
}
layer {
name: "relu5"
type: "ReLU"
bottom: "conv5"
top: "conv5"
}
#========= RCNN ============
layer {
name: "roi_pool_conv5"
type: "ROIPooling"
bottom: "conv5"
bottom: "rois"
top: "roi_pool_conv5"
roi_pooling_param {
pooled_w: 6
pooled_h: 6
spatial_scale: 0.0625 # 1/16
}
}
layer {
name: "fc6"
type: "InnerProduct"
bottom: "roi_pool_conv5"
top: "fc6"
param { lr_mult: 1.0 }
param { lr_mult: 2.0 }
inner_product_param {
num_output: 4096
}
}
layer {
name: "relu6"
type: "ReLU"
bottom: "fc6"
top: "fc6"
}
分析可见:
<1> roipooling的输入是conv5(featuremap)和rois(一系列目标框,大小不一);
<2>输出为roi_pool_conv5,即目标区域映射到featuremap的归一化特征pool_w×pool_h×nchannels,思想是将原图的目标区域映射到featuremap区域,然后featuremap区域划分成pool_w×pool_h块,每一块做max_pooling。
3. 源码解读:
// ------------------------------------------------------------------
// Fast R-CNN
// Copyright (c) 2015 Microsoft
// Licensed under The MIT License [see fast-rcnn/LICENSE for details]
// Written by Ross Girshick
// ------------------------------------------------------------------
#include <cfloat>
#include <string>
#include <utility>
#include <vector>
#include "caffe/blob.hpp"
#include "caffe/common.hpp"
#include "caffe/layer.hpp"
//#include "caffe/vision_layers.hpp"
#include "caffe/fast_rcnn_layers.hpp"
#include "caffe/proto/caffe.pb.h"
using std::max;
using std::min;
using std::floor;
using std::ceil;
#if _MSC_VER < 1800
inline double round(double x) {
return (x > 0.0) ? floor(x + 0.5) : ceil(x - 0.5);
}
#endif
namespace caffe {
template <typename Dtype>
void ROIPoolingLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
ROIPoolingParameter roi_pool_param = this->layer_param_.roi_pooling_param();
CHECK_GT(roi_pool_param.pooled_h(), 0)
<< "pooled_h must be > 0";
CHECK_GT(roi_pool_param.pooled_w(), 0)
<< "pooled_w must be > 0";
pooled_height_ = roi_pool_param.pooled_h(); //roi_pooling的宽和高
pooled_width_ = roi_pool_param.pooled_w();
spatial_scale_ = roi_pool_param.spatial_scale(); // roi_pooling的空间映射尺寸,大小为featuremap/原图
LOG(INFO) << "Spatial scale: " << spatial_scale_;
}
template <typename Dtype>
void ROIPoolingLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
channels_ = bottom[0]->channels();
height_ = bottom[0]->height();
width_ = bottom[0]->width();
top[0]->Reshape(bottom[1]->num(), channels_, pooled_height_,
pooled_width_); //输出top的尺度大小(bottom[1]_num,channels_, pooled_height_, pooled_width)
max_idx_.Reshape(bottom[1]->num(), channels_, pooled_height_,
pooled_width_);
}
template <typename Dtype>
void ROIPoolingLayer<Dtype>::Forward_cpu(const vector<Blob<Dtype>*>& bottom,
const vector<Blob<Dtype>*>& top) {
const Dtype* bottom_data = bottom[0]->cpu_data();
const Dtype* bottom_rois = bottom[1]->cpu_data();
// Number of ROIs
int num_rois = bottom[1]->num(); // 目标区域个数
int batch_size = bottom[0]->num(); // 处理图像的batch个数
int top_count = top[0]->count();
Dtype* top_data = top[0]->mutable_cpu_data();
caffe_set(top_count, Dtype(-FLT_MAX), top_data);
int* argmax_data = max_idx_.mutable_cpu_data();
caffe_set(top_count, -1, argmax_data);
// For each ROI R = [batch_index x1 y1 x2 y2]: max pool over R
for (int n = 0; n < num_rois; ++n) {//处理每一个目标区域得到归一化的maxpool特征映射
int roi_batch_ind = bottom_rois[0];
int roi_start_w = round(bottom_rois[1] * spatial_scale_);
int roi_start_h = round(bottom_rois[2] * spatial_scale_); //roi映射到featuremap区域坐标
int roi_end_w = round(bottom_rois[3] * spatial_scale_);
int roi_end_h = round(bottom_rois[4] * spatial_scale_);
CHECK_GE(roi_batch_ind, 0); //roi_batch_ind间于0到batch_size之间
CHECK_LT(roi_batch_ind, batch_size);
int roi_height = max(roi_end_h - roi_start_h + 1, 1);//映射区域宽高
int roi_width = max(roi_end_w - roi_start_w + 1, 1);
const Dtype bin_size_h = static_cast<Dtype>(roi_height)// 每一个pool块对应的高度和宽度
/ static_cast<Dtype>(pooled_height_);
const Dtype bin_size_w = static_cast<Dtype>(roi_width)
/ static_cast<Dtype>(pooled_width_);
const Dtype* batch_data = bottom_data + bottom[0]->offset(roi_batch_ind);//指向对应featuremap数据
for (int c = 0; c < channels_; ++c) {
for (int ph = 0; ph < pooled_height_; ++ph) {
for (int pw = 0; pw < pooled_width_; ++pw) {
// Compute pooling region for this output unit:
// start (included) = floor(ph * roi_height / pooled_height_)
// end (excluded) = ceil((ph + 1) * roi_height / pooled_height_)
// 做以0为起点的坐标映射
int hstart = static_cast<int>(floor(static_cast<Dtype>(ph)
* bin_size_h));
int wstart = static_cast<int>(floor(static_cast<Dtype>(pw)
* bin_size_w));
int hend = static_cast<int>(ceil(static_cast<Dtype>(ph + 1)
* bin_size_h));
int wend = static_cast<int>(ceil(static_cast<Dtype>(pw + 1)
* bin_size_w));
// 映射为目标区域的坐标映射(同时保证不越界)
hstart = min(max(hstart + roi_start_h, 0), height_);
hend = min(max(hend + roi_start_h, 0), height_);
wstart = min(max(wstart + roi_start_w, 0), width_);
wend = min(max(wend + roi_start_w, 0), width_);
//计算每一个映射特征块的最大值作为特征,有问题时赋值0和-1
bool is_empty = (hend <= hstart) || (wend <= wstart);
const int pool_index = ph * pooled_width_ + pw;
if (is_empty) {
top_data[pool_index] = 0;
argmax_data[pool_index] = -1;
}
for (int h = hstart; h < hend; ++h) {
for (int w = wstart; w < wend; ++w) {
const int index = h * width_ + w;
if (batch_data[index] > top_data[pool_index]) {
top_data[pool_index] = batch_data[index];
argmax_data[pool_index] = index;
}
}
}
}
}
// Increment all data pointers by one channel
batch_data += bottom[0]->offset(0, 1);
top_data += top[0]->offset(0, 1);
argmax_data += max_idx_.offset(0, 1);
}
// Increment ROI data pointer
bottom_rois += bottom[1]->offset(1);
}
}
template <typename Dtype>
void ROIPoolingLayer<Dtype>::Backward_cpu(const vector<Blob<Dtype>*>& top,
const vector<bool>& propagate_down, const vector<Blob<Dtype>*>& bottom) {
NOT_IMPLEMENTED;
}
#ifdef CPU_ONLY
STUB_GPU(ROIPoolingLayer);
#endif
INSTANTIATE_CLASS(ROIPoolingLayer);
REGISTER_LAYER_CLASS(ROIPooling);
} // namespace caffe