Face Detection Using KinectV2 and OpenCV - c++

Recently I have been trying to learn OpenCV and was trying to detect faces using the Haar Classifier.
I was successful in detecting faces when I obtained the video stream from the default webcam,but when I use Kinect instead of the default webcam, It still detects face but the frame rate drops tremendously.
The code that I've Written is
int main() {
string haar_face = "F:\haarcascade_frontalface_default.xml";
CascadeClassifier haar_cascade;
haar_cascade.load(haar_face);
if (haar_cascade.empty()) {
return -1;
}
vector<Rect_<int>> faces;
bool optionKinect = false;
cout << "Choose Option\n1.) Kinect \n>1.) WebCam\n";
int choice;
cin >> choice;
if (choice == 1) {
optionKinect = true;
}
if (optionKinect) {
CKinectStreamsMat* kinectStream = new CKinectStreamsMat();
kinectStream->initSensor();
while (true) {
Mat original, gray;
Mat face;
Rect face_i;
//cap >> original;
original = kinectStream->getColorFrame();
if (original.data) {
cvtColor(original, gray, CV_BGR2GRAY);
haar_cascade.detectMultiScale(gray, faces);
int size = faces.size();
for (size_t i = 0; i < size; i++) {
face_i = faces[i];
face = gray(face_i);
rectangle(original, face_i, CV_RGB(0, 255, 0), 1);
}
imshow("original", original);
if(waitKey(20) == 27){
break;
}
}
}
}
else {
VideoCapture cap(0);
while (true) {
Mat original, gray;
Mat face;
Rect face_i;
cap >> original;
//original = kinectStream->getColorFrame();
if (original.data) {
cvtColor(original, gray, CV_BGR2GRAY);
haar_cascade.detectMultiScale(gray, faces);
int size = faces.size();
for (size_t i = 0; i < size; i++) {
face_i = faces[i];
face = gray(face_i);
rectangle(original, face_i, CV_RGB(0, 255, 0), 1);
}
imshow("original", original);
if(waitKey(20) == 27){
break;
}
}
}
}
}
And this is how I am obtaining the Color frame from Kinect.
cv::Mat CKinectStreamsMat::getColorFrame()
{
HRESULT hr = E_FAIL;
IColorFrame* frame = NULL;
IFrameDescription* frameDesc;
cv::Mat colorImage;
hr = _color_reader->AcquireLatestFrame(&frame);
if (SUCCEEDED(hr)) {
hr = frame->get_FrameDescription(&frameDesc);
if (SUCCEEDED(hr)) {
int frameWidth = 0, frameHeight = 0;
hr = frameDesc->get_Width(&frameWidth);
if (SUCCEEDED(hr)) {
hr = frameDesc->get_Height(&frameHeight);
}
if (SUCCEEDED(hr)) {
const int imgSize = frameWidth*frameHeight * 4 * sizeof(unsigned char); //4 Channels(BGRA)
colorImage = cv::Mat(1080,1920,CV_8UC4);
hr = frame->CopyConvertedFrameDataToArray(imgSize, reinterpret_cast<BYTE*>(colorImage.data), ColorImageFormat_Bgra);
}
}
SafeRelease(frameDesc);
SafeRelease(frame);
}
return colorImage;
}
I thought the reason for low performance might be the difference in resolution of the frames provided by the WebCam and Kinect, So I also tried scaling down the frame provided by Kinect down to the size that was even less than the frame size of WebCam. But still the performance was very low.
As this is only what I could think of and now I am out of ideas, So could anyone please tell what could be the reason for this low performance?

Related

How to display the final decision for face detection using opencv

I am a beginner of using opencv. I using opencv 2.4 with VS2013. I had develop face detection code and success but the decision made is from frame by frame. How can I made final decision by combining all the frame or frame averaging?? For example, when the total face detected is 90% detected, the final decision is 'FACE DETECTED' and vice versa.
Here is my code:
int main(int argc, char** argv)
{
CvCapture* capture;
capture = cvCaptureFromFile("C:/Users/user/Desktop/Face and Motion/45.avi");
//assert(capture != NULL); //terminate if capture is NULL
IplImage* frame;
while (true)
{
frame = cvQueryFrame(capture);
if (!frame)
break;
cvShowImage("original", frame); //show
CvMemStorage* storage = cvCreateMemStorage(0);
CvHaarClassifierCascade* cascade = (CvHaarClassifierCascade*)cvLoad("C:/opencv2410/sources/data/haarcascades/haarcascade_frontalface_alt.xml");
cvClearMemStorage(storage);
CvSeq* faces = cvHaarDetectObjects(frame, cascade, storage, 1.1, 4, 0, cvSize(40, 50));
CvRect* r;
if (faces) //change from (!faces) to (faces)
{
for (int i = 0; i < (faces ? faces->total : 0); i++)
{
r = (CvRect*)cvGetSeqElem(faces, i);
//cvRectangle(frame, cvPoint(100, 50), cvPoint(200, 200), CV_RGB(255, 0, 0), 5, 8);
cvRectangle(frame, cvPoint(r->x, r->y), cvPoint(r->x + r->width, r->y + r->height), CV_RGB(255, 0, 0));
cvShowImage("RESULTS", frame);
char c = cvWaitKey(1000);
}
}
else
{
cvShowImage("RESULT", frame);
}
printf("%d Face Found !\n", faces ? faces->total : 0);
if ((faces ? faces->total : 0) == 0)
{
printf("FACE NOT DETECTED !\n\n");
}
else
{
printf("FACE DETECTED !\n\n");
}
}
return (0);
}
you need two variables like
int frame_count = 0;
int detected_face_count = 0;
you should increase frame_count for every frame and increase detected_face_count when a face detected. finally detected_face_count / frame_count gives your desired value.
however, i did some corrections on your code (i could not test it)
changes explained by comments. i hope it will be useful.
int main(int argc, char** argv)
{
CvCapture* capture;
capture = cvCaptureFromFile("C:/Users/user/Desktop/Face and Motion/45.avi");
//assert(capture != NULL); //terminate if capture is NULL
IplImage* frame;
//you need two variables
int frame_count = 0;
int detected_face_count = 0;
// these lines should not be in the loop
CvMemStorage* storage = cvCreateMemStorage(0);
CvHaarClassifierCascade* cascade = (CvHaarClassifierCascade*)cvLoad("C:/opencv2410/sources/data/haarcascades/haarcascade_frontalface_alt.xml");
while (true)
{
frame = cvQueryFrame(capture);
if (!frame)
break;
frame_count++; // increase frame_count
cvShowImage("original", frame); //show
cvClearMemStorage(storage);
CvSeq* faces = cvHaarDetectObjects(frame, cascade, storage, 1.1, 4, 0, cvSize(40, 50));
CvRect* r;
for (int i = 0; i < (faces ? faces->total : 0); i++)
{
r = (CvRect*)cvGetSeqElem(faces, i);
//cvRectangle(frame, cvPoint(100, 50), cvPoint(200, 200), CV_RGB(255, 0, 0), 5, 8);
cvRectangle(frame, cvPoint(r->x, r->y), cvPoint(r->x + r->width, r->y + r->height), CV_RGB(255, 0, 0));
}
cvShowImage("RESULT", frame);
char c = cvWaitKey(1000);
printf("%d Face Found !\n", faces ? faces->total : 0);
if ((faces ? faces->total : 0) == 0)
{
printf("FACE NOT DETECTED !\n\n");
}
else
{
detected_face_count ++; // increase detected_face_count
printf("FACE DETECTED !\n\n");
}
}
if( (float)(detected_face_count / frame_count) > 0.89 )
{
printf("FACES DETECTED over %90 of frames!\n\n");
}
return (0);
}
as #Miki stated you should not use deprecated C api
take a look at the code below using C++ api (tested with OpenCV 3.1)
#include <opencv2/opencv.hpp>
using namespace cv;
using namespace std;
int main(int argc, char** argv)
{
VideoCapture capture("C:/Users/user/Desktop/Face and Motion/45.avi");
if(!capture.isOpened())
{
printf("Video file could not be opened !\n");
return -1;
}
Mat frame;
//you need two variables
int frame_count = 0;
int detected_face_count = 0;
// these lines should not be in the loop
CascadeClassifier cascade;
if( !cascade.load( "C:/opencv2410/sources/data/haarcascades/haarcascade_frontalface_alt.xml" ) )
{
cerr << "ERROR: Could not load classifier cascade" << endl;
return -1;
}
while (true)
{
capture >>frame;
if (frame.empty())
break;
frame_count++; // increase frame_count
imshow("original", frame); //show
vector<Rect> faces;
cascade.detectMultiScale(frame, faces,1.1, 4, 0, Size(40, 50));
for (int i = 0; i < faces.size(); i++)
{
rectangle(frame, faces[i], Scalar(0, 0, 255));
}
imshow("RESULT", frame);
char c = waitKey(10);
printf("%d Face Found !\n", faces.size());
if ( faces.size() == 0 )
{
printf("FACE NOT DETECTED !\n\n");
}
else
{
detected_face_count ++; // increase detected_face_count
printf("FACE DETECTED !\n\n");
}
}
printf("count of frames : %d \n", frame_count);
printf("count of frames has detected face : %d \n", detected_face_count);
printf("Face Found %d percent of frames!\n", (int)(100 * detected_face_count / frame_count));
if( (float)(detected_face_count / frame_count) > 0.89 )
{
printf("FACES DETECTED over %90 of frames!\n\n");
}
return (0);
}

Store the centroid to a new array or vector

This is the image
This is the code below
#include "opencv2/opencv.hpp"
using namespace cv;
using namespace std;
int main(int argc, char** argv)
{
// Read image
Mat im = imread("beethoven_ode_to_joy.jpg", IMREAD_GRAYSCALE);
// Setup SimpleBlobDetector parameters.
SimpleBlobDetector::Params params;
// Change thresholds
params.minThreshold = 10;
params.maxThreshold = 200;
// Filter by Area.
params.filterByArea = true;
params.minArea = 15;
// Filter by Circularity
params.filterByCircularity = true;
params.minCircularity = 0.1;
// Filter by Convexity
params.filterByConvexity = true;
params.minConvexity = 0.01;
// Filter by Inertia
params.filterByInertia = true;
params.minInertiaRatio = 0.01;
// Storage for blobs
vector<KeyPoint> keypoints;
// Set up detector with params
SimpleBlobDetector detector(params);
// Detect blobs
detector.detect(im, keypoints);
At this portion of the code is printing all the centroid of the blob, but i need to store the centroid of the blob into a new vector
for (vector<KeyPoint>::iterator it = keypoints.begin(); it != keypoints.end(); ++it)
{
KeyPoint k = *it;
cout << k.pt << endl;
}
// Draw detected blobs as red circles.
// DrawMatchesFlags::DRAW_RICH_KEYPOINTS flag ensures
// the size of the circle corresponds to the size of blob
Mat im_with_keypoints;
drawKeypoints(im, keypoints, im_with_keypoints, Scalar(0, 0, 255), DrawMatchesFlags::DRAW_RICH_KEYPOINTS);
// Show blobs
imshow("keypoints", im_with_keypoints);
imwrite("a.jpg", im_with_keypoints);
waitKey(0);
}
Instead of:
for (vector<KeyPoint>::iterator it = keypoints.begin(); it != keypoints.end(); ++it)
{
KeyPoint k = *it;
cout << k.pt << endl;
}
You can do this:
std::vector<decltype(keypoints[0].pt)> vector_of_points;
keypoints.reserve(keypoints.size());
for (vector<KeyPoint>::iterator it = keypoints.begin(); it != keypoints.end(); ++it)
{
KeyPoint k = *it;
cout << k.pt << endl;
vector_of_points.emplace_back(k.pt);
}
Or even better:
std::vector<decltype(keypoints[0].pt)> vector_of_points(keypoints.size());
std::transform(std::begin(keypoints),std::end(keypoints),std::begin(vector_of_points),[](KeyPoint const& item){std::cout<<item.pt;return item.pt;});
EDIT:
If you want to access the centriod you can do this:
for(auto& centr:vector_of_points){
//centr is the centriod
}
If you want a specific centroid. The fourth for example:
vector_of_points[4]

opencv playing video slow speed

I have a problem with video speed, when I just run a video file in opencv it plays in normal speed.
The problem starts when I apply face and eye detection to the video it plays very slowly. I have tried to change waitKey() values , but the problem still exists.
Why is this happening?
#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;
// Function Headers
void detectAndDisplay(Mat frame);
// Global variables
// Copy this file from opencv/data/haarscascades to target folder
string face_cascade_name = "haarcascade_frontalface_alt.xml";
string eye_cascade_name = "haarcascade_mcs_eyepair_big.xml";
CascadeClassifier face_cascade;
CascadeClassifier eye_cascade;
string window_name = "Capture - Face detection";
int filenumber; // Number of file to be saved
string filename;
int main(void)
{
VideoCapture capture("m.mp4");
if (!capture.isOpened()) // check if we succeeded
return -1;
// Load the cascade
if (!face_cascade.load(face_cascade_name))
{
printf("--(!)Error loading\n");
return (-1);
}
if (!eye_cascade.load(eye_cascade_name))
{
printf("--(!)Error loading\n eye ");
return (-1);
}
// Read the video stream
Mat frame;
for (;;)
{
capture >> frame;
// Apply the classifier to the frame
if (!frame.empty())
{
detectAndDisplay(frame);
}
else
{
printf(" --(!) No captured frame -- Break!");
break;
}
waitKey(40);
return 0;
}
}
void detectAndDisplay(Mat frame)
{
std::vector<Rect> faces;
std::vector<Rect> eyes;
Mat frame_gray;
Mat crop;
Mat res;
Mat gray;
string text;
stringstream sstm;
cvtColor(frame, frame_gray, COLOR_BGR2GRAY);
equalizeHist(frame_gray, frame_gray);
// Detect faces
face_cascade.detectMultiScale(frame_gray, faces, 1.1, 4, CASCADE_SCALE_IMAGE, Size(20, 20));
// Set Region of Interest
size_t i = 0; // ic is index of current element
for (i = 0; i < faces.size(); i++) // Iterate through all current elements (detected faces)
{
Point pt1(faces[i].x, faces[i].y); // Display detected faces on main window - live stream from camera
Point pt2((faces[i].x + faces[i].height), (faces[i].y + faces[i].width));
rectangle(frame, pt1, pt2, Scalar(0, 255, 0), 2, 8, 0);
// set ROI for the eyes
Rect Roi = faces[i];
Roi.height = Roi.height / 4;
Roi.y = Roi.y + Roi.height;
cv::Mat crop = frame(Roi);
cvtColor(crop, gray, CV_BGR2GRAY);
imshow("ROI", gray);
eye_cascade.detectMultiScale(gray, eyes, 1.1, 2, 0 | CV_HAAR_SCALE_IMAGE, Size(20, 20));
for (size_t j = 0; j < eyes.size(); j++)
{
Rect r(Roi.x + eyes[j].x, Roi.y + eyes[j].y, eyes[j].width, eyes[j].height);
rectangle(frame, r, Scalar(255, 0, 0), 3, 1, 0);
// convert roi to ychannel
Mat eye_region = frame(eyes[j]).clone();
Mat eye_region_YCbCr;
cvtColor(eye_region, eye_region_YCbCr, CV_BGR2YCrCb);
vector<Mat> channels;
split(eye_region_YCbCr, channels);
Mat eye_region_Ychannel = channels[0].clone();
imshow("Y", eye_region_Ychannel);
// end of convert
}
}
// Show image
imshow("original", frame);
if (!crop.empty())
{
imshow("detected", crop);
}
else
{
destroyWindow("detected");
}
}

Saving Images of Detected Faces in OpenCV

I have a code that detects faces and saves multiple cropped area images of them to a file path. My code doesn't stop saving images of detected faces until I physically close the program. For every one second a face is detected on a webcam, my code saves 6 images of the face.
Is it possible to have it save just one image per face detected? For example, if there is one face, only one image, if two faces, an image of both faces are saved etc. My code is below. Can anyone help me?
#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;
void detectAndDisplay(Mat frame);
string face_cascade_name = "C:\\opencv\\sources\\data\\haarcascades\\haarcascade_frontalface_alt2.xml";
CascadeClassifier face_cascade;
string window_name = "Window";
int filenumber;
string filename;
int main(void)
{
VideoCapture capture(0);
if (!capture.isOpened())
return -1;
if (!face_cascade.load(face_cascade_name))
{
cout << "error" << endl;
return (-1);
};
Mat frame;
for (;;)
{
capture >> frame;
if (!frame.empty())
{
detectAndDisplay(frame);
}
else
{
cout << "error2" << endl;
break;
}
int c = waitKey(10);
if (27 == char(c))
{
break;
}
}
return 0;
}
void detectAndDisplay(Mat frame)
{
std::vector<Rect> faces;
Mat frame_gray;
Mat crop;
Mat res;
Mat gray;
string text;
stringstream sstm;
cvtColor(frame, frame_gray, COLOR_BGR2GRAY);
equalizeHist(frame_gray, frame_gray);
face_cascade.detectMultiScale(frame_gray, faces, 1.1, 2, 0 | CASCADE_SCALE_IMAGE, Size(30, 30));
cv::Rect roi_b;
cv::Rect roi_c;
size_t ic = 0;
int ac = 0;
size_t ib = 0;
int ab = 0;
for (ic = 0; ic < faces.size(); ic++)
{
roi_c.x = faces[ic].x;
roi_c.y = faces[ic].y;
roi_c.width = (faces[ic].width);
roi_c.height = (faces[ic].height);
ac = roi_c.width * roi_c.height;
roi_b.x = faces[ib].x;
roi_b.y = faces[ib].y;
roi_b.width = (faces[ib].width);
roi_b.height = (faces[ib].height);
crop = frame(roi_b);
resize(crop, res, Size(128, 128), 0, 0, INTER_LINEAR);
cvtColor(crop, gray, CV_BGR2GRAY);
filename = "C:\\Users\\Desktop\\Faces\\face";
stringstream ssfn;
ssfn << filename.c_str() << filenumber << ".jpg";
filename = ssfn.str();
cv::imwrite(filename, res);
filenumber++;
Point pt1(faces[ic].x, faces[ic].y);
Point pt2((faces[ic].x + faces[ic].height), (faces[ic].y + faces[ic].width));
rectangle(frame, pt1, pt2, Scalar(0, 255, 0), 2, 8, 0);
}
sstm << "Crop area size: " << roi_b.width << "x" << roi_b.height << " Filename: " << filename;
text = sstm.str();
if (!crop.empty())
{
imshow("detected", crop);
}
else
destroyWindow("detected");
}

How do i actually set optical flow on half of the image captured?

I already had an optical flow code implemented using C++ in OpenCV. However, i would like to detect optical flow in half of the image frame. Which part should i edit? is it from this function below?
cvCalcOpticalFlowPyrLK(
frame1_1C, frame2_1C,
pyramid1, pyramid2,
frame1_features,
frame2_features,
number_of_features,
optical_flow_window,
5,
optical_flow_found_feature,
optical_flow_feature_error,
optical_flow_termination_criteria,
0 );
No. There are no changes necessary in the function itself. All you need to do is pass only the part of image on which you want to calculate optical flow to the function.
You can define the range of the image that you want to carry out the optical flow calculations on. using
wanted_image=image(Range(x1,y1), Range(x2,y2))
The following is a working code based on the lkdemo.cpp in the samples folder. THe only worthwhile change is
gray = gray(Range(1,480), Range(1,320));
//Gives the left half of the image
which defines the region of interest.
#include "opencv2/video/tracking.hpp"
#include "opencv2/imgproc/imgproc.hpp"
#include "opencv2/highgui/highgui.hpp"
#include <iostream>
#include <ctype.h>
using namespace cv;
using namespace std;
static void help()
{
cout << "*** Using OpenCV version " << CV_VERSION <<" ***"<< endl;
cout << "\n\nUsage: \n"
"\tESC - quit the program\n"
"\tr - auto-initialize tracking\n"
"\tc - delete all the points\n"
"\tn - switch the \"night\" mode on/off\n"<< endl;
}
int main( int argc, char** argv )
{
help();
//Termination of the algo after 20 iterations or accuracy going under 0.03
TermCriteria termcrit(CV_TERMCRIT_ITER|CV_TERMCRIT_EPS, 20, 0.3);
Size subPixWinSize(10,10), winSize(31,31);
const int MAX_COUNT = 500;
bool needToInit = false;
bool nightMode = false;
//Video capture is from the default device i.e. the webcam
VideoCapture cap(0);
if( !cap.isOpened() )
{
cout << "Could not initialize capturing...\n";
return 0;
}
namedWindow( "Half screen Optical flow Demo!", 1 );
Mat gray, prevGray, image;
vector<Point2f> points[2];
for(;;)
{
Mat frame;
//Output from the Videocapture is piped to 'frame'
cap >> frame;
if( frame.empty() )
break;
frame.copyTo(image);
cvtColor(image, gray, COLOR_BGR2GRAY);
// Night mode not disabled
if( nightMode )
image = Scalar::all(0);
gray = gray(Range(1,480), Range(1,320));
if( needToInit || points[0].size()<=5)
{
goodFeaturesToTrack(gray, points[1], MAX_COUNT, 0.01, 10, Mat(), 3, 0, 0.4);
cornerSubPix(gray, points[1], subPixWinSize, Size(-1,-1), termcrit);
}
else if( !points[0].empty() )
{
vector<uchar> status;
vector<float> err;
if(prevGray.empty())
gray.copyTo(prevGray);
calcOpticalFlowPyrLK(prevGray, gray, points[0], points[1], status, err, winSize, 3, termcrit, 0, 0.001);
size_t i, k;
for( i = k = 0; i < points[1].size(); i++ )
{
if( !status[i] )
continue;
points[1][k++] = points[1][i];
circle(image, points[1][i], 3, Scalar(0,255,0), -1, 8);
}
points[1].resize(k);
}
needToInit = false;
imshow("Half screen Optical flow Demo!", image);
char c = (char)waitKey(10);
if( c == 27 )
break;
switch( c )
{
case 'r':
needToInit = true;
break;
case 'c':
points[0].clear();
points[1].clear();
break;
case 'n':
nightMode = !nightMode;
break;
}
std::swap(points[1], points[0]);
cv::swap(prevGray, gray);
}
cap.release();
return 0;
}
if you want to detect optical flow only in the half of the image, then you can simply give halves of the images (frame1_1C, frame2_1C) as parameters. For example, following code initializes a matrix belonging to the left half of frame1_1C:
cv::Mat frame1_1C_half(frame1_1C, cv::Range(0, frame1_1C.rows), cv::Range(0, frame1_1C.cols/2));