读trainCascade的训练结果的代码笔记
这个是OpenCV提供的人脸检测调用trainCascade训练结果的代码。最近一直在用,不过才发现里面用到的接口不是trainCascade里的,用的是modules\objdetect\里的objdetect.hpp & cascadedetect.cpp。
#include "opencv2/objdetect/objdetect.hpp"
#include "opencv2/highgui/highgui.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include <iostream>
#include <stdio.h>
using namespace std;
using namespace cv;
/** 函数声明 */
void detectAndDisplay( Mat frame );
/** 全局变量 */
string face_cascade_name = "haarcascade_frontalface_alt.xml";
string eyes_cascade_name = "haarcascade_eye_tree_eyeglasses.xml";
CascadeClassifier face_cascade;
CascadeClassifier eyes_cascade;
string window_name = "Capture - Face detection";
RNG rng(12345);
/** @主函数 */
int main( int argc, const char** argv )
{
CvCapture* capture;
Mat frame;
//-- 1. 加载级联分类器文件
if( !face_cascade.load( face_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
if( !eyes_cascade.load( eyes_cascade_name ) ){ printf("--(!)Error loading\n"); return -1; };
//-- 2. 打开内置摄像头视频流
capture = cvCaptureFromCAM( -1 );
if( capture )
{
while( true )
{
frame = cvQueryFrame( capture );
//-- 3. 对当前帧使用分类器进行检测
if( !frame.empty() )
{ detectAndDisplay( frame ); }
else
{ printf(" --(!) No captured frame -- Break!"); break; }
int c = waitKey(10);
if( (char)c == 'c' ) { break; }
}
}
return 0;
}
/** @函数 detectAndDisplay */
void detectAndDisplay( Mat frame )
{
std::vector<Rect> faces;
Mat frame_gray;
cvtColor( frame, frame_gray, CV_BGR2GRAY );
equalizeHist( frame_gray, frame_gray );
//-- 多尺寸检测人脸
face_cascade.detectMultiScale( frame_gray, faces, 1.1, 2, 0|CV_HAAR_SCALE_IMAGE, Size(30, 30) );
for( int i = 0; i < faces.size(); i++ )
{
Point center( faces[i].x + faces[i].width*0.5, faces[i].y + faces[i].height*0.5 );
ellipse( frame, center, Size( faces[i].width*0.5, faces[i].height*0.5), 0, 0, 360, Scalar( 255, 0, 255 ), 4, 8, 0 );
Mat faceROI = frame_gray( faces[i] );
std::vector<Rect> eyes;
//-- 在每张人脸上检测双眼
eyes_cascade.detectMultiScale( faceROI, eyes, 1.1, 2, 0 |CV_HAAR_SCALE_IMAGE, Size(30, 30) );
for( int j = 0; j < eyes.size(); j++ )
{
Point center( faces[i].x + eyes[j].x + eyes[j].width*0.5, faces[i].y + eyes[j].y + eyes[j].height*0.5 );
int radius = cvRound( (eyes[j].width + eyes[i].height)*0.25 );
circle( frame, center, radius, Scalar( 255, 0, 0 ), 4, 8, 0 );
}
}
//-- 显示结果图像
imshow( window_name, frame );
}
里面用到的load和detectMultiScale函数都是OpenCV的CascadeClassifier类的内容,而不是训练的时候用的CvCascadeClassifier类。以load为例:
bool CascadeClassifier::load(const string& filename)
{
oldCascade.release();
data = Data();
featureEvaluator.release();
FileStorage fs(filename, FileStorage::READ);
if( !fs.isOpened() )
return false;
if( read(fs.getFirstTopLevelNode()) )//开始读xml文件
return true;
fs.release();
oldCascade = Ptr<CvHaarClassifierCascade>((CvHaarClassifierCascade*)cvLoad(filename.c_str(), 0, 0, 0));//旧版本的xml文件
return !oldCascade.empty();
}
bool CascadeClassifier::read(const FileNode& root)
{
if( !data.read(root) )//Step 1:这里应该是读xml开头的params以及stage里的信息
return false;
// load features
featureEvaluator = FeatureEvaluator::create(data.featureType);
FileNode fn = root[CC_FEATURES];//Step 2:这里开始应该是读xml下面那块<features>开始的rect
if( fn.empty() )
return false;
return featureEvaluator->read(fn);
}
这里step 1 和 2 就是之前分析xml格式的时候的信息了。
//Step 1:
bool CascadeClassifier::Data::read(const FileNode &root)
{
static const float THRESHOLD_EPS = 1e-5f;
// load stage params
string stageTypeStr = (string)root[CC_STAGE_TYPE];
if( stageTypeStr == CC_BOOST )
stageType = BOOST;
else
return false;
string featureTypeStr = (string)root[CC_FEATURE_TYPE];
if( featureTypeStr == CC_HAAR )
featureType = FeatureEvaluator::HAAR;
else if( featureTypeStr == CC_LBP )
featureType = FeatureEvaluator::LBP;
else if( featureTypeStr == CC_HOG )
featureType = FeatureEvaluator::HOG;
else
return false;
origWinSize.width = (int)root[CC_WIDTH];
origWinSize.height = (int)root[CC_HEIGHT];
CV_Assert( origWinSize.height > 0 && origWinSize.width > 0 );
isStumpBased = (int)(root[CC_STAGE_PARAMS][CC_MAX_DEPTH]) == 1 ? true : false;
// load feature params
FileNode fn = root[CC_FEATURE_PARAMS];
if( fn.empty() )
return false;
ncategories = fn[CC_MAX_CAT_COUNT];
int subsetSize = (ncategories + 31)/32,
nodeStep = 3 + ( ncategories>0 ? subsetSize : 1 );
// load stages
fn = root[CC_STAGES];
if( fn.empty() )
return false;
stages.reserve(fn.size());
classifiers.clear();
nodes.clear();
FileNodeIterator it = fn.begin(), it_end = fn.end();
for( int si = 0; it != it_end; si++, ++it )
{
FileNode fns = *it;
Stage stage;
stage.threshold = (float)fns[CC_STAGE_THRESHOLD] - THRESHOLD_EPS;
fns = fns[CC_WEAK_CLASSIFIERS];
if(fns.empty())
return false;
stage.ntrees = (int)fns.size();
stage.first = (int)classifiers.size();
stages.push_back(stage);
classifiers.reserve(stages[si].first + stages[si].ntrees);
FileNodeIterator it1 = fns.begin(), it1_end = fns.end();
for( ; it1 != it1_end; ++it1 ) // weak trees
{
FileNode fnw = *it1;
FileNode internalNodes = fnw[CC_INTERNAL_NODES];//internal nodes
FileNode leafValues = fnw[CC_LEAF_VALUES];//leaf value
if( internalNodes.empty() || leafValues.empty() )
return false;
DTree tree;
tree.nodeCount = (int)internalNodes.size()/nodeStep;
classifiers.push_back(tree);
nodes.reserve(nodes.size() + tree.nodeCount);
leaves.reserve(leaves.size() + leafValues.size());
if( subsetSize > 0 )
subsets.reserve(subsets.size() + tree.nodeCount*subsetSize);
FileNodeIterator internalNodesIter = internalNodes.begin(), internalNodesEnd = internalNodes.end();
for( ; internalNodesIter != internalNodesEnd; ) // nodes
{
DTreeNode node;
node.left = (int)*internalNodesIter; ++internalNodesIter;
node.right = (int)*internalNodesIter; ++internalNodesIter;
node.featureIdx = (int)*internalNodesIter; ++internalNodesIter;
if( subsetSize > 0 )
{
for( int j = 0; j < subsetSize; j++, ++internalNodesIter )
subsets.push_back((int)*internalNodesIter);
node.threshold = 0.f;
}
else
{
node.threshold = (float)*internalNodesIter; ++internalNodesIter;
}
nodes.push_back(node);
}
internalNodesIter = leafValues.begin(), internalNodesEnd = leafValues.end();
for( ; internalNodesIter != internalNodesEnd; ++internalNodesIter ) // leaves
leaves.push_back((float)*internalNodesIter);
}
}
return true;
}
//Step 2:以LBP为例
bool LBPEvaluator::read( const FileNode& node )
{
features->resize(node.size());
featuresPtr = &(*features)[0];
FileNodeIterator it = node.begin(), it_end = node.end();
for(int i = 0; it != it_end; ++it, i++)
{
if(!featuresPtr[i].read(*it))
return false;
}
return true;
}
bool LBPEvaluator::Feature :: read(const FileNode& node )
{
FileNode rnode = node[CC_RECT];
FileNodeIterator it = rnode.begin();
it >> rect.x >> rect.y >> rect.width >> rect.height;
return true;
}
My Q:至今不太明白为啥有trainCascade和CvtrainCascade,FeatureEvaluator和CvFeatureEvaluator等等,怎么感觉很多是重复的东西呢。
补充:
关于FeatureEvaluator 和 CvFeatureEvaluator的关系:
区别来看的话:
1. 首先FeatureEvaluator是在modules\objdetect\下,而CvFeatureEvaluator定义在apps\trainCascade\下。
2. FeatureEvaluator应该是去调用训练好的结果。而CvFeatureEvaluator应该是用来去训练而设计的。因为在FeatureEvaluator里只有read, 没有write。相反在CvFeatureEvaluator中只有write没有read。再有就是FeatureEvaluator里没有关于class Label和sample index的参数接口,而CvFeatureEvaluator里面有。
3. CvFeatureEvaluator把有些部分分出去作为Params
两者是有很多共同的地方。比如计算图像的积分图,创建函数create( int featureType )。因为这部分不管是训练还是调用训练后的结果都要用到。