JNIEXPORT jint JNICALL NAME(NativeCalculate)(JNIEnv * jenv, jclass s, jstring jPicName)
{
if(mLastretValue !=ERR_NONE)
return mLastretValue;
const char* strImgName = jenv->GetStringUTFChars(jPicName, NULL);
int retValue = ERR_NONE;
try
{
msg_Err("#---strImgName=%s",strImgName);
// Read Image
cv::Mat garyImgData = cv::imread(strImgName, CV_LOAD_IMAGE_GRAYSCALE);
cv::Mat oriImgData = cv::imread(strImgName, CV_LOAD_IMAGE_COLOR);
msg_Err("#---444444,garyImgData.cols,rows=%d,%d",garyImgData.cols,garyImgData.rows);
// Face detection
int ndFaceNum = 100;
smart_ptr<FaceRect> pFaceRects(100);
int nRetCode = FaceDetect(garyImgData.data, garyImgData.cols, garyImgData.rows,
24, std::min(garyImgData.cols, garyImgData.rows), pFaceRects, &ndFaceNum);
if (0 >= ndFaceNum || ERR_NONE != nRetCode)
{
FaceDecetedCallback(jenv,JNI_FALSE);
if(0 >= ndFaceNum && nRetCode==ERR_NONE)
nRetCode = -666;
msg_Err("#---Can not detect any faces!");
std::cout << "Can not detect any faces!" << std::endl;
throw nRetCode;
}
FaceDecetedCallback(jenv,JNI_TRUE);
msg_Err("#---55555");
// Pick up the largest one as the true face
int faceSize = -1;
int faceIdx = 0;
for (int i = 0; i < ndFaceNum; ++i)
{
if (ndFaceNum > 1 && pFaceRects[i].confidence < 0.45f)
continue;
if (faceSize < pFaceRects[i].right - pFaceRects[i].left)
{
faceSize = pFaceRects[i].right - pFaceRects[i].left;
faceIdx = i;
}
}
msg_Err("#---66666");
// Face aligment
smart_ptr<FeaPointF> FeaPoints(FEATURE_POINTS_NUMBER);
nRetCode = FaceAlignment(garyImgData.data, garyImgData.cols, garyImgData.rows, pFaceRects + faceIdx, FeaPoints);
if (ERR_NONE != nRetCode)
{
std::cout << "Failed to locate feature points on faces!" << std::endl;
throw nRetCode;
}
// Choose eye corner and lip corner, suppose1cdhmopruw
// The normalization method should be considered by some days
// Use old normalization method
smart_ptr<FeaPointF> NormPoint(FEATURE_POINTS_NUMBER);
memset(NormPoint, 0, FEATURE_POINTS_NUMBER * sizeof(FeaPointF));
NormPoint[8].x = 70; NormPoint[8].y = 100;
NormPoint[10].x = 120; NormPoint[10].y = 100;
NormPoint[11].x = 180; NormPoint[11].y = 100;
NormPoint[9].x = 230; NormPoint[9].y = 100;
NormPoint[22].x = 110; NormPoint[22].y = 200;
NormPoint[23].x = 190; NormPoint[23].y = 200;
double weight[88] = { 0 };
weight[8] = 1;
weight[10] = 1;
weight[11] = 1;
weight[9] = 1;
weight[22] = 1;
weight[23] = 1;
smart_ptr<double> pDstImage(300 * 300 * oriImgData.channels());
nRetCode = NormalizeFace(oriImgData.data, oriImgData.cols, oriImgData.rows, oriImgData.channels(), FeaPoints, NormPoint,
weight, FEATURE_POINTS_NUMBER, 300, 300, pDstImage);
if (ERR_NONE != nRetCode)
{
std::cout << "Failed to normalize faces!" << std::endl;
throw nRetCode;
}
msg_Err("#---77777");
smart_ptr<unsigned char> pNormFace(300 * 300 * 3);
for (int p = 0; p < 300 * 300 * 3; ++p)
pNormFace[p] = (unsigned char)(int(pDstImage[p]));
smart_ptr<unsigned char> pCropNormFace(256 * 256 * 3);
smart_ptr<unsigned char> pCropGrayNormFace(256 * 256);
cv::Mat NormFaceImage(300, 300, CV_8UC3, pNormFace);
cv::Rect roi(22, 22, 256, 256);
cv::Mat CropImage = NormFaceImage(roi);
cv::Mat GrayCropImage;
cv::cvtColor(CropImage, GrayCropImage, cv::COLOR_BGR2GRAY);
for (int h = 0; h < CropImage.rows; ++h) {
const uchar* ptr = CropImage.ptr<uchar>(h);
int img_index = 0;
for (int w = 0; w < CropImage.cols; ++w) {
for (int c = 0; c < CropImage.channels(); ++c) {
int datum_index = (c * CropImage.rows + h) * CropImage.cols + w;
pCropNormFace[datum_index] = static_cast<char>(ptr[img_index++]);
}
}
}
for (int h = 0; h < GrayCropImage.rows; ++h) {
const uchar* ptr = GrayCropImage.ptr<uchar>(h);
int img_index = 0;
for (int w = 0; w < GrayCropImage.cols; ++w) {
for (int c = 0; c < GrayCropImage.channels(); ++c) {
int datum_index = (c * GrayCropImage.rows + h) * GrayCropImage.cols + w;
pCropGrayNormFace[datum_index] = static_cast<char>(ptr[img_index++]);
}
}
}
msg_Err("#---888888");
// 1、总体分
int featDim = GetDeepFeatSize(hFace) / 4;
smart_ptr<float> pFeatures(featDim);
nRetCode = InnerDeepFeat(hFace, pCropNormFace, 1, 3, 256, 256, pFeatures);
float score = pFeatures[0] * 1.11f;
if (score > 100.0f)
score = 100.0f;
std::cout << "Total score: " << score << std::endl;
mTotalScore = score;
msg_Err("#---99999");
// 2、瑕疵
featDim = GetDeepFeatSize(hXiaci) / 4;
pFeatures.reset(featDim);
nRetCode = InnerDeepFeat(hXiaci, pCropNormFace, 1, 3, 256, 256, pFeatures);
float maxR = -10000.0f;
int label = 15;
for (int j = 0; j < featDim; ++j)
{
//std::cout << pFeatures[j] << " ";
if (maxR < pFeatures[j])
{
maxR = pFeatures[j];
label = j;
}
}
if (0 == label)
std::cout << "The flaws' number: " << "none!" << std::endl;
else if (1 == label)
std::cout << "The flaws' number: " << "a little!" << std::endl;
else if (2 == label)
std::cout << "The flaws' number: " << "small!" << std::endl;
else if (3 == label)
std::cout << "The flaws' number: " << "a lot!" << std::endl;
else if (4 == label)
std::cout << "The flaws' number: " << "very much!" << std::endl;
mFlawLabel = label;
msg_Err("#---aaaaa");
// 3、开心
featDim = GetDeepFeatSize(hHappy) / 4;
pFeatures.reset(featDim);
nRetCode = InnerDeepFeat(hHappy, pCropNormFace, 1, 3, 256, 256, pFeatures);
maxR = -10000.0f;
label = 15;
for (int j = 0; j < featDim; ++j)
{
//std::cout << pFeatures[j] << " ";
if (maxR < pFeatures[j])
{
maxR = pFeatures[j];
label = j;
}
}
// std::cout << std::endl;
if (0 == label)
std::cout << "Angry!" << std::endl;
else if (1 == label)
std::cout << "Unhappy!" << std::endl;
else if (2 == label)
std::cout << "normal!" << std::endl;
else if (3 == label)
std::cout << "happy!" << std::endl;
else if (4 == label)
std::cout << "smile!" << std::endl;
mExpressionLabel = label;
msg_Err("#---bbbbbb");
// 4、年龄
smart_ptr<unsigned char> pDstImageII(128*128);
nRetCode = NormliazeFaceII(oriImgData.data, oriImgData.cols, oriImgData.rows, FeaPoints, FEATURE_POINTS_NUMBER, pDstImageII);
featDim = GetDeepFeatSize(hAge) / 4;
pFeatures.reset(featDim);
nRetCode = InnerDeepFeat(hAge, pDstImageII, 1, 1, 128, 128, pFeatures);
score = pFeatures[0];
std::cout << "Age: " << score << std::endl;
mAge = score;
msg_Err("#---ccccc");
// 5、肤色
featDim = GetDeepFeatSize(hSkin) / 4;
pFeatures.reset(featDim);
nRetCode = InnerDeepFeat(hSkin, pCropNormFace, 1, 3, 256, 256, pFeatures);
score = pFeatures[0] * 1.11f;
if (score > 100.0f)
score = 100.0f;
std::cout << "Skin score: " << score << std::endl;
mSkin = score;
}
catch (const std::bad_alloc &)
{
retValue = ERR_MEMORYALLOC;
}
catch (const int &errCode)
{
retValue = errCode;
}
catch (...)
{
retValue = ERR_UNKNOWN;
}
jenv->ReleaseStringUTFChars(jPicName,strImgName);
return retValue;
}
int FaceDetectWrapperF64(const double *img, int width, int height, const char
*classifier, int *faces, int maxdetect) {
return FaceDetect(img,width,height,classifier,faces,maxdetect);
}
