ssd网络详解之priorbox layer

ssd 网络详解之priorbox layer

本文原创,转载请注明出处:https://blog.csdn.net/dan_teng/article/details/81532013

ssd网络一大特点是,为了提高检测准确率,在不同尺度的特征图上进行预测,这种预测就需要prior box layer。
prior box 是干嘛的呢?其实非常类似于Faster R-CNN中的Anchors,就是候选框,这种候选框的选取不需要像R-CNN那样通过复杂处理产生。在ssd中,priorbox层只需要bottom层feature map的大小,就可以给出候选框。假设输入的feature map大小是W×H,生成的prior box中心就有W×H个,均匀分布在整张图上。在每个中心上,可以生成多个不同长宽比的prior box,如[1/3, 1/2, 1, 2, 3],每个点就可以生成length_of_aspect_ratio个框,所以在一个feature map上可以生成的prior box总数是W×H×length_of_aspect_ratio。
ssd网络详解之priorbox layer_第1张图片
如上图b)所示,在8x8的feature map上,每个中心点生成了4个预选框。(c)是4x4的feature map,假设(c)是(b)后面的层,prior box参数相同,那么显然,从(b)中提取的预选框更有利于小尺度对象的检出,而从(c)中提取的预选框更有利于大尺度对象的检出(4x4的feature map 只有8x8 feature map长宽的1/2,相同参数的prior box圈出的面积相同,但是两个feature map对应于相同的原图,因此4x4 feature map上相同面积映射回原图相当于圈出了更大的面积,是8x8面积的4倍)。
如上所述,SSD提取了不同尺度的feature map来做检测,大尺度特征图(较靠前的特征图)可以用来检测小物体,而小尺度特征图(较靠后的特征图)用来检测大物体。从下图可以看出,ssd网络会从6层feature map中提取特征,priorbox相应也会跟在这6层后面。
ssd网络详解之priorbox layer_第2张图片
计算思路:以feature map上每个点的中点为中心(offset=0.5),生成一些列同心的prior box(然后中心点的坐标会乘以step,相当于从feature map位置映射回原图位置),最后会归一化处理
正方形prior box最小边长为min_size,最大边长为

minsizemaxsize m i n s i z e ∗ m a x s i z e

每在prototxt设置一个aspect ratio,会生成2个长方形,长宽为:
aspect_ratiominsize a s p e c t _ r a t i o ∗ m i n s i z e
1/aspect_ratiominsize 1 / a s p e c t _ r a t i o ∗ m i n s i z e

ssd网络详解之priorbox layer_第3张图片
之前看过别人博客中对源码的注释,有一些含糊之处。由于本人采用不同的架构重新实现了一遍,对每个细节都要理解透,因此在下面的代码中,尽量给出详细的注释。对于默认值的解释,在代码之后放上了caffe中的定义,感兴趣的可以到文末参阅。

#include 
#include 
#include 
#include 

#include "caffe/layers/prior_box_layer.hpp"

namespace caffe {

template <typename Dtype>
void PriorBoxLayer::LayerSetUp(const vector*>& bottom,
      const vector*>& top) {
  const PriorBoxParameter& prior_box_param =
      this->layer_param_.prior_box_param();
  CHECK_GT(prior_box_param.min_size_size(), 0) << "must provide min_size.";
  // 取min_size数据
  for (int i = 0; i < prior_box_param.min_size_size(); ++i) {
    min_sizes_.push_back(prior_box_param.min_size(i));
    CHECK_GT(min_sizes_.back(), 0) << "min_size must be positive.";
  }
  aspect_ratios_.clear();
  aspect_ratios_.push_back(1.);//1是默认比例
  flip_ = prior_box_param.flip();
  //存放aspect_ratios数据
  for (int i = 0; i < prior_box_param.aspect_ratio_size(); ++i) {
    float ar = prior_box_param.aspect_ratio(i);
    bool already_exist = false;
    for (int j = 0; j < aspect_ratios_.size(); ++j) {
      if (fabs(ar - aspect_ratios_[j]) < 1e-6) {
        already_exist = true;
        break;
      }
    }
    if (!already_exist) {
      aspect_ratios_.push_back(ar);
      if (flip_) {//是否取倒数
        aspect_ratios_.push_back(1./ar);
      }
    }
  }
  // prior box 数量。注意:还需加上max_size数量才是最终数量
  num_priors_ = aspect_ratios_.size() * min_sizes_.size();
  if (prior_box_param.max_size_size() > 0) {
    CHECK_EQ(prior_box_param.min_size_size(), prior_box_param.max_size_size());
    for (int i = 0; i < prior_box_param.max_size_size(); ++i) {
      max_sizes_.push_back(prior_box_param.max_size(i));
      CHECK_GT(max_sizes_[i], min_sizes_[i])
          << "max_size must be greater than min_size.";
      num_priors_ += 1;// 这里增加num_priors数量
    }
  }
  clip_ = prior_box_param.clip();
  // variance与后期真实框计算有关,要么给1个值,要么给4个值
  if (prior_box_param.variance_size() > 1) {
    // Must and only provide 4 variance.
    CHECK_EQ(prior_box_param.variance_size(), 4);
    for (int i = 0; i < prior_box_param.variance_size(); ++i) {
      CHECK_GT(prior_box_param.variance(i), 0);
      variance_.push_back(prior_box_param.variance(i));
    }
  } else if (prior_box_param.variance_size() == 1) {
    CHECK_GT(prior_box_param.variance(0), 0);
    variance_.push_back(prior_box_param.variance(0));
  } else {
    // Set default to 0.1.
    variance_.push_back(0.1);
  }

  if (prior_box_param.has_img_h() || prior_box_param.has_img_w()) {
    CHECK(!prior_box_param.has_img_size())
        << "Either img_size or img_h/img_w should be specified; not both.";
    img_h_ = prior_box_param.img_h();
    CHECK_GT(img_h_, 0) << "img_h should be larger than 0.";
    img_w_ = prior_box_param.img_w();
    CHECK_GT(img_w_, 0) << "img_w should be larger than 0.";
  } else if (prior_box_param.has_img_size()) {
    const int img_size = prior_box_param.img_size();
    CHECK_GT(img_size, 0) << "img_size should be larger than 0.";
    img_h_ = img_size;
    img_w_ = img_size;
  } else {
    img_h_ = 0;
    img_w_ = 0;
  }

  if (prior_box_param.has_step_h() || prior_box_param.has_step_w()) {
    CHECK(!prior_box_param.has_step())
        << "Either step or step_h/step_w should be specified; not both.";
    step_h_ = prior_box_param.step_h();
    CHECK_GT(step_h_, 0.) << "step_h should be larger than 0.";
    step_w_ = prior_box_param.step_w();
    CHECK_GT(step_w_, 0.) << "step_w should be larger than 0.";
  } else if (prior_box_param.has_step()) {
    const float step = prior_box_param.step();
    CHECK_GT(step, 0) << "step should be larger than 0.";
    step_h_ = step;
    step_w_ = step;
  } else {
    step_h_ = 0;
    step_w_ = 0;
  }

  offset_ = prior_box_param.offset();
  reduce_boxes_ = prior_box_param.reduce_boxes();
}
//该层输出大小为【1,2,layer_width * layer_height * num_priors_ * 4】
// c的第一维,存放每个框的四个点
// c的第二维,存放variance(每个框都一样)
template <typename Dtype>
void PriorBoxLayer::Reshape(const vector*>& bottom,
      const vector*>& top) {
  // 取feature map大小
  const int layer_width = bottom[0]->width();
  const int layer_height = bottom[0]->height();
  vector<int> top_shape(3, 1);
  // Since all images in a batch has same height and width, we only need to
  // generate one set of priors which can be shared across all images.
  top_shape[0] = 1;
  // 2 channels. First channel stores the mean of each prior coordinate.
  // Second channel stores the variance of each prior coordinate.
  top_shape[1] = 2;
  top_shape[2] = layer_width * layer_height * num_priors_ * 4;
  CHECK_GT(top_shape[2], 0);
  top[0]->Reshape(top_shape);
}

template <typename Dtype>
void PriorBoxLayer::Forward_cpu(const vector*>& bottom,
    const vector*>& top) {
  // 取feature map大小
  const int layer_width = bottom[0]->width();
  const int layer_height = bottom[0]->height();
  int img_width, img_height;
  if (img_h_ == 0 || img_w_ == 0) {
    img_width = bottom[1]->width();// 输入图像的宽高
    img_height = bottom[1]->height();
  } else {
    img_width = img_w_;
    img_height = img_h_;
  }
  float step_w, step_h;
  if (step_w_ == 0 || step_h_ == 0) {
    step_w = static_cast<float>(img_width) / layer_width;// 缩放比例
    step_h = static_cast<float>(img_height) / layer_height;
  } else {
    step_w = step_w_;
    step_h = step_h_;
  }
  Dtype* top_data = top[0]->mutable_cpu_data();
  // 最后一维输出大小
  int dim = layer_height * layer_width * num_priors_ * 4;
  int idx = 0;
  for (int h = 0; h < layer_height; ++h) {
    for (int w = 0; w < layer_width; ++w) {
      // 取feature map 每个点为中心点,进行处理
      // offset默认值是0.5,可理解为一个小的偏移量
      // 这里将中心点映射回了原图
      float center_x = (w + offset_) * step_w;
      float center_y = (h + offset_) * step_h;
      float box_width, box_height;
      for (int s = 0; s < min_sizes_.size(); ++s) {
        int min_size_ = min_sizes_[s];
        // first prior: aspect_ratio = 1, size = min_size

        if (reduce_boxes_) {
          box_width = box_height = min_size_ / 2.0;// for mobilenet, conv11 featuremap
        }
        else{
          box_width = box_height = min_size_;
        }
        // min_size确定的正方形框,大小进行了归一化
        // xmin
        top_data[idx++] = (center_x - box_width / 2.) / img_width;
        // ymin
        top_data[idx++] = (center_y - box_height / 2.) / img_height;
        // xmax
        top_data[idx++] = (center_x + box_width / 2.) / img_width;
        // ymax
        top_data[idx++] = (center_y + box_height / 2.) / img_height;

        if (max_sizes_.size() > 0) {
          CHECK_EQ(min_sizes_.size(), max_sizes_.size());
          int max_size_ = max_sizes_[s];
          // second prior: aspect_ratio = 1, size = sqrt(min_size * max_size)
          box_width = box_height = sqrt(min_size_ * max_size_);
          // max_size确定的正方形框
          // xmin
          top_data[idx++] = (center_x - box_width / 2.) / img_width;
          // ymin
          top_data[idx++] = (center_y - box_height / 2.) / img_height;
          // xmax
          top_data[idx++] = (center_x + box_width / 2.) / img_width;
          // ymax
          top_data[idx++] = (center_y + box_height / 2.) / img_height;
        }

        // rest of priors
        for (int r = 0; r < aspect_ratios_.size(); ++r) {
          float ar = aspect_ratios_[r];
          if (fabs(ar - 1.) < 1e-6) {
            continue;
          }
          // 根据定义,由aspect_ratio和min_size共同确定的矩形框
          box_width = min_size_ * sqrt(ar);
          box_height = min_size_ / sqrt(ar);
          // xmin
          top_data[idx++] = (center_x - box_width / 2.) / img_width;
          // ymin
          top_data[idx++] = (center_y - box_height / 2.) / img_height;
          // xmax
          top_data[idx++] = (center_x + box_width / 2.) / img_width;
          // ymax
          top_data[idx++] = (center_y + box_height / 2.) / img_height;
        }
      }
    }
  }
  // clip默认值是false,是否进行越界处理
  // clip the prior's coordidate such that it is within [0, 1]
  if (clip_) {
    for (int d = 0; d < dim; ++d) {
      top_data[d] = std::min(std::max(top_data[d], 0.), 1.);
    }
  }
  // 前面提到过,输出c维大小是2,第一部分存放预选框数据,第二部分存放variance
  // set the variance.
  top_data += top[0]->offset(0, 1);// 通过偏移拿到第二部分的地址
  if (variance_.size() == 1) {
    caffe_set(dim, Dtype(variance_[0]), top_data);
  } else {
    int count = 0;
    for (int h = 0; h < layer_height; ++h) {
      for (int w = 0; w < layer_width; ++w) {
        for (int i = 0; i < num_priors_; ++i) {
          for (int j = 0; j < 4; ++j) {
            top_data[count] = variance_[j];
            ++count;
          }
        }
      }
    }
  }
}

INSTANTIATE_CLASS(PriorBoxLayer);
REGISTER_LAYER_CLASS(PriorBox);

}  // namespace caffe

caffe中的定义:

// Message that store parameters used by PriorBoxLayer
message PriorBoxParameter {
  // Encode/decode type.
  enum CodeType {// 编码方式,与训练有关,且与最后一层detection out解码有关
    CORNER = 1;
    CENTER_SIZE = 2;
    CORNER_SIZE = 3;
  }
  // Minimum box size (in pixels). Required!
  repeated float min_size = 1;// 最小尺寸
  // Maximum box size (in pixels). Required!
  repeated float max_size = 2;// 最大尺寸
  // Various of aspect ratios. Duplicate ratios will be ignored.
  // If none is provided, we use default ratio 1.
  repeated float aspect_ratio = 3;// 变换比例
  // If true, will flip each aspect ratio.
  // For example, if there is aspect ratio "r",
  // we will generate aspect ratio "1.0/r" as well.
  optional bool flip = 4 [default = true]; // 对每个变换比例是否再加上他们的倒数
  // If true, will clip the prior so that it is within [0, 1]
  optional bool clip = 5 [default = false]; // 是否裁剪到[0,1]
  // Variance for adjusting the prior bboxes.
  repeated float variance = 6;
  // By default, we calculate img_height, img_width, step_x, step_y based on
  // bottom[0] (feat) and bottom[1] (img). Unless these values are explicitely
  // provided.
  // Explicitly provide the img_size.
  optional uint32 img_size = 7;
  // Either img_size or img_h/img_w should be specified; not both.
  optional uint32 img_h = 8;
  optional uint32 img_w = 9;

  // Explicitly provide the step size.
  optional float step = 10;
  // Either step or step_h/step_w should be specified; not both.
  optional float step_h = 11;
  optional float step_w = 12;

  // Offset to the top left corner of each cell.
  optional float offset = 13 [default = 0.5]; // 偏移量

  optional bool reduce_boxes = 14 [default = false];
}

你可能感兴趣的:(cnn,SSD,检测)