// Get potential bounding boxes for all test images
void Objectness::getObjBndBoxesForTests(vector<vector<Vec4i>> &_boxesTests, int numDetPerSize)
{
const int TestNum = _voc.testSet.size();
vecM imgs3u(TestNum);
vector<ValStructVec<float, Vec4i>> boxesTests;
boxesTests.resize(TestNum);
#pragma omp parallel for
for (int i = 0; i < TestNum; i++){
//imgs3u[i] = imread(format(_S(_voc.imgPathW), _S(_voc.testSet[i])));
imgs3u[i] = imread("C:/Users/TerryChen/Desktop/M.jpg");
boxesTests[i].reserve(10000);
}
int scales[3] = {1, 3, 5};
for (int clr = MAXBGR; clr <= G; clr++){
setColorSpace(clr);
trainObjectness(numDetPerSize);
loadTrainedModel();
CmTimer tm("Predict");
tm.Start();
#pragma omp parallel for
for (int i = 0; i < TestNum; i++){
ValStructVec<float, Vec4i> boxes;
getObjBndBoxes(imgs3u[i], boxes, numDetPerSize);
boxesTests[i].append(boxes, scales[clr]);
}
tm.Stop();
printf("Average time for predicting an image (%s) is %gs\n", _clrName[_Clr], tm.TimeInSeconds()/TestNum);
}
_boxesTests.resize(TestNum);
CmFile::MkDir(_bbResDir);
#pragma omp parallel for
for (int i = 0; i < TestNum; i++){
CStr fName = _bbResDir + _voc.testSet[i];
ValStructVec<float, Vec4i> &boxes = boxesTests[i];
FILE *f = fopen(_S(fName + ".txt"), "w");
fprintf(f, "%d\n", boxes.size());
for (size_t k = 0; k < boxes.size(); k++)
fprintf(f, "%g, %s\n", boxes(k), _S(strVec4i(boxes[k])));
fclose(f);
_boxesTests[i].resize(boxesTests[i].size());
for (int j = 0; j < boxesTests[i].size(); j++)
_boxesTests[i][j] = boxesTests[i][j];
}
evaluatePerImgRecall(_boxesTests, "PerImgAllNS.m", 5000);
#pragma omp parallel for
for (int i = 0; i < TestNum; i++){
boxesTests[i].sort(false);
for (int j = 0; j < boxesTests[i].size(); j++)
_boxesTests[i][j] = boxesTests[i][j];
}
evaluatePerImgRecall(_boxesTests, "PerImgAllS.m", 5000);
}
// Get potential bounding boxes for all test images
void Objectness::getObjBndBoxesForTestsFast(vector<vector<Vec4i> > &_boxesTests, int numDetPerSize, bool preloadModel, bool preloadImages)
{
//setColorSpace(HSV);
if (!preloadModel)
trainObjectness(numDetPerSize);
loadTrainedModel();
illustrate();
const int TestNum = _voc.testSet.size();
vecM imgs3u;
if (preloadImages)
imgs3u.resize(TestNum);
vector<ValStructVec<float, Vec4i> > boxesTests;
boxesTests.resize(TestNum);
//Reading all images beforehand needs a lot of memory.
//The timing results are affected based on when the image is loaded.
if (preloadImages) {
#pragma omp parallel for
for (int i = 0; i < TestNum; i++) {
imgs3u[i] = imread(format(_S(_voc.imgPathW), _S(_voc.testSet[i])));
boxesTests[i].reserve(10000);
}
}
printf("Start predicting\n");
CmTimer tm("Predict");
tm.Start();
#pragma omp parallel for
for (int i = 0; i < TestNum; i++) {
if (!preloadImages) {
Mat img = imread(format(_S(_voc.imgPathW), _S(_voc.testSet[i])));
getObjBndBoxes(img, boxesTests[i], numDetPerSize);
} else {
getObjBndBoxes(imgs3u[i], boxesTests[i], numDetPerSize);
}
}
tm.Stop();
if (!preloadImages) {
printf("Average time for loading and predicting an image (%s) is %gs\n", _clrName[_Clr], tm.TimeInSeconds()/TestNum);
} else {
printf("Average time for predicting an image (%s) is %gs\n", _clrName[_Clr], tm.TimeInSeconds()/TestNum);
}
_boxesTests.resize(TestNum);
CmFile::MkDir(_bbResDir);
#pragma omp parallel for
for (int i = 0; i < TestNum; i++) {
CStr fName = _bbResDir + _voc.testSet[i];
ValStructVec<float, Vec4i> &boxes = boxesTests[i];
FILE *f = fopen(_S(fName + ".txt"), "w");
fprintf(f, "%d\n", boxes.size());
for (size_t k = 0; k < boxes.size(); k++)
fprintf(f, "%g, %s\n", boxes(k), _S(strVec4i(boxes[k])));
fclose(f);
_boxesTests[i].resize(boxesTests[i].size());
for (int j = 0; j < boxesTests[i].size(); j++)
_boxesTests[i][j] = boxesTests[i][j];
}
evaluatePerImgRecall(_boxesTests, "PerImgAll.m", 5000);
}
void Objectness::trainStateII(int numPerSz)
{
loadTrainedModel();
const int NUM_TRAIN = _voc.trainNum;
vector<vecI> SZ(NUM_TRAIN), Y(NUM_TRAIN);
vector<vecF> VAL(NUM_TRAIN);
#pragma omp parallel for
for (int i = 0; i < _voc.trainNum; i++) {
const vector<Vec4i> &bbgts = _voc.gtTrainBoxes[i];
ValStructVec<float, Vec4i> valBoxes;
vecI &sz = SZ[i], &y = Y[i];
vecF &val = VAL[i];
CStr imgPath = format(_S(_voc.imgPathW), _S(_voc.trainSet[i]));
predictBBoxSI(imread(imgPath), valBoxes, sz, numPerSz, false);
const int num = valBoxes.size();
CV_Assert(sz.size() == num);
y.resize(num), val.resize(num);
for (int j = 0; j < num; j++) {
Vec4i bb = valBoxes[j];
val[j] = valBoxes(j);
y[j] = maxIntUnion(bb, bbgts) >= 0.5 ? 1 : -1;
}
}
const int NUM_SZ = _svmSzIdxs.size();
const int maxTrainNum = 100000;
vector<vecM> rXP(NUM_SZ), rXN(NUM_SZ);
for (int r = 0; r < NUM_SZ; r++) {
rXP[r].reserve(maxTrainNum);
rXN[r].reserve(1000000);
}
for (int i = 0; i < NUM_TRAIN; i++) {
const vecI &sz = SZ[i], &y = Y[i];
vecF &val = VAL[i];
int num = sz.size();
for (int j = 0; j < num; j++) {
int r = sz[j];
CV_Assert(r >= 0 && r < NUM_SZ);
if (y[j] == 1)
rXP[r].push_back(Mat(1, 1, CV_32F, &val[j]));
else
rXN[r].push_back(Mat(1, 1, CV_32F, &val[j]));
}
}
Mat wMat(NUM_SZ, 2, CV_32F);
for (int i = 0; i < NUM_SZ; i++) {
const vecM &xP = rXP[i], &xN = rXN[i];
if (xP.size() < 10 || xN.size() < 10)
printf("Warning %s:%d not enough training sample for r[%d] = %d. P = %d, N = %d\n", __FILE__, __LINE__, i, _svmSzIdxs[i], xP.size(), xN.size());
for (size_t k = 0; k < xP.size(); k++)
CV_Assert(xP[k].size() == Size(1, 1) && xP[k].type() == CV_32F);
Mat wr = trainSVM(xP, xN, L1R_L2LOSS_SVC, 100, 1);
CV_Assert(wr.size() == Size(2, 1));
wr.copyTo(wMat.row(i));
}
matWrite(_modelName + ".wS2", wMat);
_svmReW1f = wMat;
}
