double-hit-balls/game/main_code.cpp

562 lines
18 KiB
C++
Executable File

#include "main_code.h"
#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include <time.h>
#include "include/Engine.h"
#include "main_code.h"
//#define USE_PREDICTION
cv::Point flipVertical(cv::Point point) {
return cv::Point(point.x, 720 - point.y);
}
FaceStruct::FaceStruct()
{
modelPoints.push_back(cv::Point3d(6.825897, 6.760612, 4.402142)); // left brow left corner
modelPoints.push_back(cv::Point3d(1.330353, 7.122144, 6.903745)); // left brow right corner
modelPoints.push_back(cv::Point3d(-1.330353, 7.122144, 6.903745)); // right brow left corner
modelPoints.push_back(cv::Point3d(-6.825897, 6.760612, 4.402142)); // right brow right corner
modelPoints.push_back(cv::Point3d(5.311432, 5.485328, 3.987654)); // left eye left corner
modelPoints.push_back(cv::Point3d(1.789930, 5.393625, 4.413414)); // left eye right corner
modelPoints.push_back(cv::Point3d(-1.789930, 5.393625, 4.413414)); // right eye left corner
modelPoints.push_back(cv::Point3d(-5.311432, 5.485328, 3.987654)); // right eye right corner
modelPoints.push_back(cv::Point3d(2.005628, 1.409845, 6.165652)); // nose left corner
modelPoints.push_back(cv::Point3d(-2.005628, 1.409845, 6.165652)); // nose right corner
modelPoints.push_back(cv::Point3d(2.774015, -2.080775, 5.048531)); // mouth left corner
modelPoints.push_back(cv::Point3d(-2.774015, -2.080775, 5.048531)); // mouth right corner
modelPoints.push_back(cv::Point3d(0.000000, -3.116408, 6.097667)); // mouth central bottom corner
modelPoints.push_back(cv::Point3d(0.000000, -7.415691, 4.070434)); // chin corner
poseMatrix = cv::Mat(3, 4, CV_64FC1);
eulerAngles = cv::Mat(3, 1, CV_64FC1);
boxModelSrc.push_back(cv::Point3d(10.0, 10.0, 10.0));
boxModelSrc.push_back(cv::Point3d(10.0, 10.0, -10.0));
boxModelSrc.push_back(cv::Point3d(10.0, -10.0, -10.0));
boxModelSrc.push_back(cv::Point3d(10.0, -10.0, 10.0));
boxModelSrc.push_back(cv::Point3d(-10.0, 10.0, 10.0));
boxModelSrc.push_back(cv::Point3d(-10.0, 10.0, -10.0));
boxModelSrc.push_back(cv::Point3d(-10.0, -10.0, -10.0));
boxModelSrc.push_back(cv::Point3d(-10.0, -10.0, 10.0));
}
FaceStruct::FaceStruct(const std::array<cv::Point2f, LANDMARK_POINT_COUNT>& fromPreds, cv::Mat& cameraMatrix, cv::Mat& distortionCoefficients)
{
FaceStruct();
ApplyPreds(fromPreds, 0.0, cameraMatrix, distortionCoefficients);
}
void FaceStruct::CalcFromPreds(bool noticeablyChanged, cv::Mat& cameraMatrix, cv::Mat& distortionCoefficients)
{
for(size_t i = 0; i < LANDMARK_POINT_COUNT; i++) {
if(noticeablyChanged) {
landmarkSmoothers[i * 2].reset();
landmarkSmoothers[i * 2 + 1].reset();
} else {
preds[i] = Vector2f(
landmarkSmoothers[i * 2].responsiveAnalogReadSimple(preds[i](0)),
landmarkSmoothers[i * 2 + 1].responsiveAnalogReadSimple(preds[i](1))
);
}
}
float minX = preds[0](0);
float maxX = preds[0](0);
float minY = preds[0](1);
float maxY = preds[0](1);
float sumX = preds[0](0);
float sumY = preds[0](1);
for (size_t i = 1; i < LANDMARK_POINT_COUNT; i++)
{
if (minX > preds[i](0))
{
minX = preds[i](0);
}
if (minY > preds[i](1))
{
minY = preds[i](1);
}
if (maxX < preds[i](0))
{
maxX = preds[i](0);
}
if (maxY < preds[i](1))
{
maxY = preds[i](1);
}
sumX += preds[i](0);
sumY += preds[i](1);
}
float avgX = sumX / static_cast<float>(LANDMARK_POINT_COUNT);
float avgY = sumY / static_cast<float>(LANDMARK_POINT_COUNT);
std::vector<cv::Point2f> imagePoints;
imagePoints.push_back(cv::Point2f(preds[17](0), preds[17](1))); // #17 left brow left corner
imagePoints.push_back(cv::Point2f(preds[21](0), preds[21](1))); // #21 left brow right corner
imagePoints.push_back(cv::Point2f(preds[22](0), preds[22](1))); // #22 right brow left corner
imagePoints.push_back(cv::Point2f(preds[26](0), preds[26](1))); // #26 right brow right corner
imagePoints.push_back(cv::Point2f(preds[36](0), preds[36](1))); // #36 left eye left corner
imagePoints.push_back(cv::Point2f(preds[39](0), preds[39](1))); // #39 left eye right corner
imagePoints.push_back(cv::Point2f(preds[42](0), preds[42](1))); // #42 right eye left corner
imagePoints.push_back(cv::Point2f(preds[45](0), preds[45](1))); // #45 right eye right corner
imagePoints.push_back(cv::Point2f(preds[31](0), preds[31](1))); // #31 nose left corner
imagePoints.push_back(cv::Point2f(preds[35](0), preds[35](1))); // #35 nose right corner
imagePoints.push_back(cv::Point2f(preds[48](0), preds[48](1))); // #48 mouth left corner
imagePoints.push_back(cv::Point2f(preds[54](0), preds[54](1))); // #54 mouth right corner
imagePoints.push_back(cv::Point2f(preds[57](0), preds[57](1))); // #57 mouth central bottom corner
imagePoints.push_back(cv::Point2f(preds[8](0), preds[8](1))); // #8 chin corner
cv::solvePnP(modelPoints, imagePoints, cameraMatrix, distortionCoefficients, rotationVector, translationVector);
//if(similarity > 0.01) {
// for (size_t i = 0; i < 3; i++) {
// rotationVectorSmoothers[i].reset();
// translationVectorSmoothers[i].reset();
// }
//}
/*for(size_t i = 0; i < 3; i++) {
rotationVector.at<double>(i) = rotationVectorSmoothers[i].responsiveAnalogReadSimple(rotationVector.at<double>(i));
translationVector.at<double>(i) = translationVectorSmoothers[i].responsiveAnalogReadSimple(translationVector.at<double>(i));
}*/
cv::projectPoints(boxModelSrc, rotationVector, translationVector, cameraMatrix, distortionCoefficients, boxModelDst);
// calc Euler angles
cv::Rodrigues(rotationVector, rotationMatrix);
cv::hconcat(rotationMatrix, translationVector, poseMatrix);
cv::Mat outIntrinsics = cv::Mat(3, 3, CV_64FC1);
cv::Mat outRotation = cv::Mat(3, 3, CV_64FC1);
cv::Mat outTranslation = cv::Mat(3, 1, CV_64FC1);
cv::decomposeProjectionMatrix(poseMatrix, outIntrinsics, outRotation, outTranslation, cv::noArray(), cv::noArray(), cv::noArray(), eulerAngles);
center = {avgX, avgY};
size = {maxX- minX, maxY - minY};
}
void FaceStruct::ApplyPreds(const std::array<cv::Point2f, LANDMARK_POINT_COUNT>& fromPreds, bool noticeablyChanged, cv::Mat& cameraMatrix, cv::Mat& distortionCoefficients)
{
InnerApplyPreds(fromPreds);
CalcFromPreds(noticeablyChanged, cameraMatrix, distortionCoefficients);
}
void FaceStruct::InnerApplyPreds(const std::array<cv::Point2f, LANDMARK_POINT_COUNT>& fromPreds)
{
for (size_t i = 0; i < LANDMARK_POINT_COUNT; i++)
{
preds[i] = { fromPreds[i].x, 720 - fromPreds[i].y};
}
}
cv::VideoCapture cap;
FaceLandmarkDetector faceLandmarkDetector;
GLuint lastFrameTexture;
std::array<FaceLandmarkStruct, MAX_FACE_COUNT> faceLandmarkArr;
cv::Mat frame;
std::vector<cv::Mat> frameArr;
int currentIndex = -1;
int lastProcessedIndex = 0;
int lineShift = 0;
int prevPassedIndex = 0;
int indexStep = 0;
std::array<FaceStruct, MAX_FACE_COUNT> faceStruct;
const std::string CONST_LOADING_BACKGROUND_BLACK = "loading_background_black";
TMyApplication* Application;
void TMyApplication::InnerInit()
{
Application = this;
#ifdef TARGET_WIN32
#ifdef NDEBUG
//ST::PathToResources = "resources/";
ST::PathToResources = "../../../assets/";
#else
ST::PathToResources = "../../../assets/";
#endif
#endif
#ifdef TARGET_IOS
ST::PathToResources = "assets/";
#endif
if (Console != NULL)
{
*Console<<"APP INIT\n";
}
srand (static_cast<size_t>(time(NULL)));
ResourceManager->ShaderManager.AddShader("DefaultShader", "shader1vertex.txt", "shader1fragment.txt");
ResourceManager->ShaderManager.AddShader("FrameShader", "frameshader_vertex.txt", "frameshader_fragment.txt");
Renderer->PushShader("DefaultShader");
ResourceManager->TexList.AddTexture(CONST_LOADING_BACKGROUND_BLACK + ".png", CONST_LOADING_BACKGROUND_BLACK);
auto texid = ResourceManager->TexList.AddTexture("emoji/Asset 3@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 4@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 5@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 6@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 7@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 8@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 9@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 10@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 11@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 13@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 14@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 15@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 16@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 17@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 18@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 19@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 20@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 21@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 22@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 23@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 24@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 25@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 26@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 27@20x.png");
ResourceManager->TexList.AddTexture("emoji/Asset 28@20x.png");
ResourceManager->FontManager.AddFont("droid_sans14", "droid_sans14_font_bitmap.png", "droid_sans14_font_charmap.txt");
ResourceManager->FontManager.PushFont("droid_sans14");
lastFrameTexture = ResourceManager->TexList.AddEmptyTexture("lastFrameTexture", 1280, 720);
Renderer->SetOrthoProjection();
Renderer->SetFullScreenViewport();
for (size_t i = 0; i < MAX_FACE_COUNT; i++)
{
faceRenderPair[i].first.SamplerMap[CONST_STRING_TEXTURE_UNIFORM] = "Asset "+boost::lexical_cast<std::string>(i+3)+"@20x.png";
faceRenderPair[i].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB].push_back(Vector3f(0, 0, 0));
faceRenderPair[i].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB].push_back(Vector3f(0, 512, 0));
faceRenderPair[i].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB].push_back(Vector3f(512, 512, 0));
faceRenderPair[i].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB].push_back(Vector3f(512, 512, 0));
faceRenderPair[i].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB].push_back(Vector3f(512, 0, 0));
faceRenderPair[i].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB].push_back(Vector3f(0, 0, 0));
faceRenderPair[i].second.Data.Vec2CoordArr[CONST_STRING_TEXCOORD_ATTRIB].push_back(Vector2f(0, 0));
faceRenderPair[i].second.Data.Vec2CoordArr[CONST_STRING_TEXCOORD_ATTRIB].push_back(Vector2f(0, 1));
faceRenderPair[i].second.Data.Vec2CoordArr[CONST_STRING_TEXCOORD_ATTRIB].push_back(Vector2f(1, 1));
faceRenderPair[i].second.Data.Vec2CoordArr[CONST_STRING_TEXCOORD_ATTRIB].push_back(Vector2f(1, 1));
faceRenderPair[i].second.Data.Vec2CoordArr[CONST_STRING_TEXCOORD_ATTRIB].push_back(Vector2f(1, 0));
faceRenderPair[i].second.Data.Vec2CoordArr[CONST_STRING_TEXCOORD_ATTRIB].push_back(Vector2f(0, 0));
faceRenderPair[i].second.RefreshBuffer();
}
cap = cv::VideoCapture(0);
//cap = cv::VideoCapture("video.mp4");
//cap = cv::VideoCapture("bp.mp4");
// Check if camera opened successfully
if (!cap.isOpened()) {
std::cout << "Error opening video stream or file" << std::endl;
}
cap.set(CV_CAP_PROP_FRAME_WIDTH, 1280);
cap.set(CV_CAP_PROP_FRAME_HEIGHT, 720);
faceLandmarkDetector.StartTrackProcess();
Inited = true;
}
void TMyApplication::InnerDeinit()
{
faceLandmarkDetector.StopTrackProcess();
cap.release();
Inited = false;
if (Console != NULL)
{
*Console<<"APP DEINIT\n";
}
}
void TMyApplication::InnerOnTapDown(Vector2f p)
{
}
void TMyApplication::InnerOnTapUp(Vector2f p)
{
}
void TMyApplication::InnerOnTapUpAfterMove(Vector2f p)
{
}
void TMyApplication::InnerOnMove(Vector2f p, Vector2f shift)
{
}
void TMyApplication::OnFling(Vector2f v)
{
}
void TMyApplication::InnerDraw()
{
glDisable(GL_DEPTH_TEST);
//glBindTexture(GL_TEXTURE_2D, ResourceManager->TexList[CONST_LOADING_BACKGROUND_BLACK]);
glBindTexture(GL_TEXTURE_2D, lastFrameTexture);
Renderer->DrawRect(Vector2f(0.f, 0.f), Vector2f(Renderer->GetMatrixWidth(), Renderer->GetMatrixHeight()));
/*
for (size_t i = 0; i < MAX_FACE_COUNT; i++)
{
TRenderParamsSetter params(faceRenderPair[i].first);
Renderer->DrawTriangleList(faceRenderPair[i].second);
}*/
}
void TMyApplication::InnerUpdate(size_t dt)
{
cap >> frame; // get a new frame from camera/video or read image
if (frame.empty())
{
return;
}
cv::Mat image = frame.clone();
currentIndex++;
faceLandmarkArr = faceLandmarkDetector.GetFaceLandmarks(frame, currentIndex);
#ifdef USE_PREDICTION
frameArr.push_back(image);
while (frameArr.size() > 40)
{
lineShift++;
frameArr.erase(frameArr.begin());
}
currentIndex++;
faceLandmarkArr = faceLandmarkDetector.GetFaceLandmarks(frame, currentIndex);
bool newProcessedIndexFound = false;
int newProcessedIndex;
for (size_t i = 0; i < faceLandmarkArr.size(); i++)
{
if (faceLandmarkArr[i].valid)
{
if (!newProcessedIndexFound)
{
newProcessedIndexFound = true;
newProcessedIndex = faceLandmarkArr[i].frameIndex;
}
else
{
if (newProcessedIndex < faceLandmarkArr[i].frameIndex)
{
newProcessedIndex = faceLandmarkArr[i].frameIndex;
}
}
}
}
if (newProcessedIndexFound && newProcessedIndex > lastProcessedIndex)
{
indexStep = 0;
lastProcessedIndex = newProcessedIndex;
prevPassedIndex = currentIndex;
}
else
{
if (prevPassedIndex > lastProcessedIndex + indexStep)
{
indexStep++;
}
}
//std::cout << currentIndex << " " << lastProcessedIndex << " stepped:" << lastProcessedIndex + indexStep << " indexStep: " << indexStep << std::endl;
cv::Mat renderImage;
if (lastProcessedIndex + indexStep - lineShift < 0)
{
renderImage = frameArr[0];
}
else if (lastProcessedIndex + indexStep - lineShift >= frameArr.size())
{
renderImage = frameArr[frameArr.size() - 1];
}
else
{
renderImage = frameArr[lastProcessedIndex + indexStep - lineShift];
}
renderImage = renderImage.clone();
#else
cv::Mat renderImage = image;
#endif
cv::Mat cameraMatrix = (cv::Mat_<float>(3, 3) << 9.5434798118121421e+02, 0.0, 640.0, 0.0, 9.5434798118121421e+02, 360.0, 0.0, 0.0, 1.0);
cv::Mat distortionCoefficients = (cv::Mat_<float>(5, 1) << 3.5475494652047600e-02, - 5.9818881494543774e-01, 0.0, 0.0, 1.4642154802454528e+00);
for (size_t i = 0; i < faceLandmarkArr.size(); i++)
{
if (faceLandmarkArr[i].valid)
{
cv::rectangle(renderImage, faceLandmarkArr[i].faceRect, cv::Scalar(0, 0, 255), 1, 4, 0);
for (size_t j = 0; j < faceLandmarkArr[i].landmarkArr.size(); j++)
{
//cv::circle(renderImage, faceLandmarkArr[i].landmarkArr[j], 0.1, cv::Scalar(0, 255, 255), 4, 8, 0);
}
}
}
for (size_t index = 0; index < faceLandmarkArr.size(); index++)
{
if (faceLandmarkArr[index].valid)
{
faceStruct[index].ApplyPreds(faceLandmarkArr[index].landmarkArr, faceLandmarkArr[index].noticeablyChanged, cameraMatrix, distortionCoefficients);
for (size_t i = 0; i < LANDMARK_POINT_COUNT; i++)
{
cv::ellipse(renderImage, cv::Point2f(faceStruct[index].preds[i](0), 720 - faceStruct[index].preds[i](1)), cv::Size(2, 2), 0, 0, 360, cv::Scalar(255, 128, 128), 4, 8, 0);
}
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[0]), flipVertical(faceStruct[index].boxModelDst[1]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[1]), flipVertical(faceStruct[index].boxModelDst[2]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[2]), flipVertical(faceStruct[index].boxModelDst[3]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[3]), flipVertical(faceStruct[index].boxModelDst[0]), cv::Scalar(0, 255, 0));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[4]), flipVertical(faceStruct[index].boxModelDst[5]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[5]), flipVertical(faceStruct[index].boxModelDst[6]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[6]), flipVertical(faceStruct[index].boxModelDst[7]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[7]), flipVertical(faceStruct[index].boxModelDst[4]), cv::Scalar(0, 255, 0));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[0]), flipVertical(faceStruct[index].boxModelDst[4]), cv::Scalar(0, 255, 0));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[1]), flipVertical(faceStruct[index].boxModelDst[5]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[2]), flipVertical(faceStruct[index].boxModelDst[6]), cv::Scalar(0, 0, 255));
cv::line(renderImage, flipVertical(faceStruct[index].boxModelDst[3]), flipVertical(faceStruct[index].boxModelDst[7]), cv::Scalar(0, 255, 0));
Vector2f posFrom = faceStruct[index].center - faceStruct[index].size*0.5f* 1.35f;
Vector2f posTo = faceStruct[index].center + faceStruct[index].size*0.5f* 1.35f;
ApplyVertexCoordVec(faceRenderPair[index].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB], posFrom, posTo, 0);
faceRenderPair[index].second.RefreshBuffer();
}
else
{
ApplyVertexCoordVec(faceRenderPair[index].second.Data.Vec3CoordArr[CONST_STRING_POSITION_ATTRIB], { 0,0 }, { 0,0 }, 0);
faceRenderPair[index].second.RefreshBuffer();
}
}
glBindTexture(GL_TEXTURE_2D, lastFrameTexture);
static std::array<char, 1280 * 720 * 3> buffer;
int step = renderImage.step;
int height = renderImage.rows;
int width = renderImage.cols;
int channels = 3;
char * data = (char *)renderImage.data;
for (int i = 0; i < height; i++)
{
int ci = height - i - 1;
memcpy(&buffer[i*width*channels], &(data[ci*step]), width*channels);
}
glTexImage2D(
GL_TEXTURE_2D,
0,
GL_RGB,
1280,
720,
0,
GL_BGR_EXT,
GL_UNSIGNED_BYTE,
&buffer[0]);
}
bool TMyApplication::IsInited()
{
return Inited;
}