IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

samples/cpp remove unused code (#16787)

Browse Source
This commit is contained in:
Zlobin Vladimir
2023-04-06 20:59:00 +04:00
committed by GitHub
parent 4812879318
commit c7c7c4bb05
5 changed files with 19 additions and 811 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
samples/cpp/benchmark_app/infer_request_wrap.hpp
View File

@@ -26,8 +26,7 @@
typedef std::function<void(size_t id, size_t group_id, const double latency, const std::exception_ptr& ptr)>
QueueCallbackFunction;
/// @brief Wrapper class for InferenceEngine::InferRequest. Handles asynchronous callbacks and calculates execution
/// time.
/// @brief Handles asynchronous callbacks and calculates execution time
class InferReqWrap final {
public:
using Ptr = std::shared_ptr<InferReqWrap>;

683
samples/cpp/common/utils/include/samples/common.hpp
View File

@@ -146,146 +146,6 @@ inline slog::LogStream& operator<<(slog::LogStream& os, const std::map<std::stri
return os;
}
/**
* @class Color
* @brief A Color class stores channels of a given color
*/
class Color {
private:
unsigned char _r;
unsigned char _g;
unsigned char _b;
public:
/**
* A default constructor.
* @param r - value for red channel
* @param g - value for green channel
* @param b - value for blue channel
*/
Color(unsigned char r, unsigned char g, unsigned char b) : _r(r), _g(g), _b(b) {}
inline unsigned char red() {
return _r;
}
inline unsigned char blue() {
return _b;
}
inline unsigned char green() {
return _g;
}
};
// TODO : keep only one version of writeOutputBMP
/**
* @brief Writes output data to image
* @param name - image name
* @param data - output data
* @param classesNum - the number of classes
* @return false if error else true
*/
static UNUSED void writeOutputBmp(std::vector<std::vector<size_t>> data, size_t classesNum, std::ostream& outFile) {
unsigned int seed = (unsigned int)time(NULL);
// Known colors for training classes from Cityscape dataset
static std::vector<Color> colors = {
{128, 64, 128}, {232, 35, 244}, {70, 70, 70}, {156, 102, 102}, {153, 153, 190}, {153, 153, 153},
{30, 170, 250}, {0, 220, 220}, {35, 142, 107}, {152, 251, 152}, {180, 130, 70}, {60, 20, 220},
{0, 0, 255}, {142, 0, 0}, {70, 0, 0}, {100, 60, 0}, {90, 0, 0}, {230, 0, 0},
{32, 11, 119}, {0, 74, 111}, {81, 0, 81}};
while (classesNum > colors.size()) {
static std::mt19937 rng(seed);
std::uniform_int_distribution<int> dist(0, 255);
Color color(dist(rng), dist(rng), dist(rng));
colors.push_back(color);
}
unsigned char file[14] = {
'B',
'M', // magic
0,
0,
0,
0, // size in bytes
0,
0, // app data
0,
0, // app data
40 + 14,
0,
0,
0 // start of data offset
};
unsigned char info[40] = {
40, 0, 0, 0, // info hd size
0, 0, 0, 0, // width
0, 0, 0, 0, // height
1, 0, // number color planes
24, 0, // bits per pixel
0, 0, 0, 0, // compression is none
0, 0, 0, 0, // image bits size
0x13, 0x0B, 0, 0, // horz resolution in pixel / m
0x13, 0x0B, 0, 0, // vert resolution (0x03C3 = 96 dpi, 0x0B13 = 72 dpi)
0, 0, 0, 0, // #colors in palette
0, 0, 0, 0, // #important colors
};
auto height = data.size();
auto width = data.at(0).size();
OPENVINO_ASSERT(
height < (size_t)std::numeric_limits<int32_t>::max && width < (size_t)std::numeric_limits<int32_t>::max,
"File size is too big: ",
height,
" X ",
width);
int padSize = static_cast<int>(4 - (width * 3) % 4) % 4;
int sizeData = static_cast<int>(width * height * 3 + height * padSize);
int sizeAll = sizeData + sizeof(file) + sizeof(info);
file[2] = (unsigned char)(sizeAll);
file[3] = (unsigned char)(sizeAll >> 8);
file[4] = (unsigned char)(sizeAll >> 16);
file[5] = (unsigned char)(sizeAll >> 24);
info[4] = (unsigned char)(width);
info[5] = (unsigned char)(width >> 8);
info[6] = (unsigned char)(width >> 16);
info[7] = (unsigned char)(width >> 24);
int32_t negativeHeight = -(int32_t)height;
info[8] = (unsigned char)(negativeHeight);
info[9] = (unsigned char)(negativeHeight >> 8);
info[10] = (unsigned char)(negativeHeight >> 16);
info[11] = (unsigned char)(negativeHeight >> 24);
info[20] = (unsigned char)(sizeData);
info[21] = (unsigned char)(sizeData >> 8);
info[22] = (unsigned char)(sizeData >> 16);
info[23] = (unsigned char)(sizeData >> 24);
outFile.write(reinterpret_cast<char*>(file), sizeof(file));
outFile.write(reinterpret_cast<char*>(info), sizeof(info));
unsigned char pad[3] = {0, 0, 0};
for (size_t y = 0; y < height; y++) {
for (size_t x = 0; x < width; x++) {
unsigned char pixel[3];
size_t index = data.at(y).at(x);
pixel[0] = colors.at(index).red();
pixel[1] = colors.at(index).green();
pixel[2] = colors.at(index).blue();
outFile.write(reinterpret_cast<char*>(pixel), 3);
}
outFile.write(reinterpret_cast<char*>(pad), padSize);
}
}
/**
* @brief Writes output data to BMP image
* @param name - image name
@@ -396,7 +256,12 @@ static UNUSED void addRectangles(unsigned char* data,
size_t width,
std::vector<int> rectangles,
std::vector<int> classes,
int thickness = 1) {
int thickness) {
struct Color {
unsigned char red;
unsigned char green;
unsigned char blue;
};
std::vector<Color> colors = {// colors to be used for bounding boxes
{128, 64, 128}, {232, 35, 244}, {70, 70, 70}, {156, 102, 102}, {153, 153, 190},
{153, 153, 153}, {30, 170, 250}, {0, 220, 220}, {35, 142, 107}, {152, 251, 152},
@@ -451,12 +316,12 @@ static UNUSED void addRectangles(unsigned char* data,
shift_first = (y + t) * width * 3;
shift_second = (y + h - t) * width * 3;
for (int ii = x; ii < x + w + 1; ii++) {
data[shift_first + ii * 3] = colors.at(cls).red();
data[shift_first + ii * 3 + 1] = colors.at(cls).green();
data[shift_first + ii * 3 + 2] = colors.at(cls).blue();
data[shift_second + ii * 3] = colors.at(cls).red();
data[shift_second + ii * 3 + 1] = colors.at(cls).green();
data[shift_second + ii * 3 + 2] = colors.at(cls).blue();
data[shift_first + ii * 3] = colors.at(cls).red;
data[shift_first + ii * 3 + 1] = colors.at(cls).green;
data[shift_first + ii * 3 + 2] = colors.at(cls).blue;
data[shift_second + ii * 3] = colors.at(cls).red;
data[shift_second + ii * 3 + 1] = colors.at(cls).green;
data[shift_second + ii * 3 + 2] = colors.at(cls).blue;
}
}
@@ -464,510 +329,17 @@ static UNUSED void addRectangles(unsigned char* data,
shift_first = (x + t) * 3;
shift_second = (x + w - t) * 3;
for (int ii = y; ii < y + h + 1; ii++) {
data[shift_first + ii * width * 3] = colors.at(cls).red();
data[shift_first + ii * width * 3 + 1] = colors.at(cls).green();
data[shift_first + ii * width * 3 + 2] = colors.at(cls).blue();
data[shift_second + ii * width * 3] = colors.at(cls).red();
data[shift_second + ii * width * 3 + 1] = colors.at(cls).green();
data[shift_second + ii * width * 3 + 2] = colors.at(cls).blue();
data[shift_first + ii * width * 3] = colors.at(cls).red;
data[shift_first + ii * width * 3 + 1] = colors.at(cls).green;
data[shift_first + ii * width * 3 + 2] = colors.at(cls).blue;
data[shift_second + ii * width * 3] = colors.at(cls).red;
data[shift_second + ii * width * 3 + 1] = colors.at(cls).green;
data[shift_second + ii * width * 3 + 2] = colors.at(cls).blue;
}
}
}
}
/**
* Write output data to image
* \param name - image name
* \param data - output data
* \param classesNum - the number of classes
* \return false if error else true
*/
static UNUSED bool writeOutputBmp(unsigned char* data, size_t height, size_t width, std::ostream& outFile) {
unsigned char file[14] = {
'B',
'M', // magic
0,
0,
0,
0, // size in bytes
0,
0, // app data
0,
0, // app data
40 + 14,
0,
0,
0 // start of data offset
};
unsigned char info[40] = {
40, 0, 0, 0, // info hd size
0, 0, 0, 0, // width
0, 0, 0, 0, // height
1, 0, // number color planes
24, 0, // bits per pixel
0, 0, 0, 0, // compression is none
0, 0, 0, 0, // image bits size
0x13, 0x0B, 0, 0, // horz resolution in pixel / m
0x13, 0x0B, 0, 0, // vert resolution (0x03C3 = 96 dpi, 0x0B13 = 72 dpi)
0, 0, 0, 0, // #colors in palette
0, 0, 0, 0, // #important colors
};
OPENVINO_ASSERT(
height < (size_t)std::numeric_limits<int32_t>::max && width < (size_t)std::numeric_limits<int32_t>::max,
"File size is too big: ",
height,
" X ",
width);
int padSize = static_cast<int>(4 - (width * 3) % 4) % 4;
int sizeData = static_cast<int>(width * height * 3 + height * padSize);
int sizeAll = sizeData + sizeof(file) + sizeof(info);
file[2] = (unsigned char)(sizeAll);
file[3] = (unsigned char)(sizeAll >> 8);
file[4] = (unsigned char)(sizeAll >> 16);
file[5] = (unsigned char)(sizeAll >> 24);
info[4] = (unsigned char)(width);
info[5] = (unsigned char)(width >> 8);
info[6] = (unsigned char)(width >> 16);
info[7] = (unsigned char)(width >> 24);
int32_t negativeHeight = -(int32_t)height;
info[8] = (unsigned char)(negativeHeight);
info[9] = (unsigned char)(negativeHeight >> 8);
info[10] = (unsigned char)(negativeHeight >> 16);
info[11] = (unsigned char)(negativeHeight >> 24);
info[20] = (unsigned char)(sizeData);
info[21] = (unsigned char)(sizeData >> 8);
info[22] = (unsigned char)(sizeData >> 16);
info[23] = (unsigned char)(sizeData >> 24);
outFile.write(reinterpret_cast<char*>(file), sizeof(file));
outFile.write(reinterpret_cast<char*>(info), sizeof(info));
unsigned char pad[3] = {0, 0, 0};
for (size_t y = 0; y < height; y++) {
for (size_t x = 0; x < width; x++) {
unsigned char pixel[3];
pixel[0] = data[y * width * 3 + x * 3];
pixel[1] = data[y * width * 3 + x * 3 + 1];
pixel[2] = data[y * width * 3 + x * 3 + 2];
outFile.write(reinterpret_cast<char*>(pixel), 3);
}
outFile.write(reinterpret_cast<char*>(pad), padSize);
}
return true;
}
static UNUSED void printPerformanceCounts(const std::map<std::string, ov::ProfilingInfo>& performanceMap,
std::ostream& stream,
std::string deviceName,
bool bshowHeader = true) {
std::chrono::microseconds totalTime = std::chrono::microseconds::zero();
// Print performance counts
if (bshowHeader) {
stream << std::endl << "performance counts:" << std::endl << std::endl;
}
std::ios::fmtflags fmt(std::cout.flags());
for (const auto& it : performanceMap) {
std::string toPrint(it.first);
const int maxLayerName = 30;
if (it.first.length() >= maxLayerName) {
toPrint = it.first.substr(0, maxLayerName - 4);
toPrint += "...";
}
stream << std::setw(maxLayerName) << std::left << toPrint;
switch (it.second.status) {
case ov::ProfilingInfo::Status::EXECUTED:
stream << std::setw(15) << std::left << "EXECUTED";
break;
case ov::ProfilingInfo::Status::NOT_RUN:
stream << std::setw(15) << std::left << "NOT_RUN";
break;
case ov::ProfilingInfo::Status::OPTIMIZED_OUT:
stream << std::setw(15) << std::left << "OPTIMIZED_OUT";
break;
}
stream << std::setw(30) << std::left << "layerType: " + std::string(it.second.node_type) + " ";
stream << std::setw(20) << std::left << "realTime: " + std::to_string(it.second.real_time.count());
stream << std::setw(20) << std::left << "cpu: " + std::to_string(it.second.cpu_time.count());
stream << " execType: " << it.second.exec_type << std::endl;
if (it.second.real_time.count() > 0) {
totalTime += it.second.real_time;
}
}
stream << std::setw(20) << std::left << "Total time: " + std::to_string(totalTime.count()) << " microseconds"
<< std::endl;
stream << std::endl;
stream << "Full device name: " << deviceName << std::endl;
stream << std::endl;
stream.flags(fmt);
}
/**
* @brief This class represents an object that is found by an object detection net
*/
class DetectedObject {
public:
int objectType;
float xmin, xmax, ymin, ymax, prob;
bool difficult;
DetectedObject(int _objectType,
float _xmin,
float _ymin,
float _xmax,
float _ymax,
float _prob,
bool _difficult = false)
: objectType(_objectType),
xmin(_xmin),
xmax(_xmax),
ymin(_ymin),
ymax(_ymax),
prob(_prob),
difficult(_difficult) {}
DetectedObject(const DetectedObject& other) = default;
static float ioU(const DetectedObject& detectedObject1_, const DetectedObject& detectedObject2_) {
// Add small space to eliminate empty squares
float epsilon = 0; // 1e-5f;
DetectedObject detectedObject1(detectedObject1_.objectType,
(detectedObject1_.xmin - epsilon),
(detectedObject1_.ymin - epsilon),
(detectedObject1_.xmax - epsilon),
(detectedObject1_.ymax - epsilon),
detectedObject1_.prob);
DetectedObject detectedObject2(detectedObject2_.objectType,
(detectedObject2_.xmin + epsilon),
(detectedObject2_.ymin + epsilon),
(detectedObject2_.xmax),
(detectedObject2_.ymax),
detectedObject2_.prob);
if (detectedObject1.objectType != detectedObject2.objectType) {
// objects are different, so the result is 0
return 0.0f;
}
if (detectedObject1.xmax < detectedObject1.xmin)
return 0.0;
if (detectedObject1.ymax < detectedObject1.ymin)
return 0.0;
if (detectedObject2.xmax < detectedObject2.xmin)
return 0.0;
if (detectedObject2.ymax < detectedObject2.ymin)
return 0.0;
float xmin = (std::max)(detectedObject1.xmin, detectedObject2.xmin);
float ymin = (std::max)(detectedObject1.ymin, detectedObject2.ymin);
float xmax = (std::min)(detectedObject1.xmax, detectedObject2.xmax);
float ymax = (std::min)(detectedObject1.ymax, detectedObject2.ymax);
// Caffe adds 1 to every length if the box isn't normalized. So do we...
float addendum;
if (xmax > 1 || ymax > 1)
addendum = 1;
else
addendum = 0;
// intersection
float intr;
if ((xmax >= xmin) && (ymax >= ymin)) {
intr = (addendum + xmax - xmin) * (addendum + ymax - ymin);
} else {
intr = 0.0f;
}
// union
float square1 = (addendum + detectedObject1.xmax - detectedObject1.xmin) *
(addendum + detectedObject1.ymax - detectedObject1.ymin);
float square2 = (addendum + detectedObject2.xmax - detectedObject2.xmin) *
(addendum + detectedObject2.ymax - detectedObject2.ymin);
float unn = square1 + square2 - intr;
return static_cast<float>(intr) / unn;
}
DetectedObject scale(float scale_x, float scale_y) const {
return DetectedObject(objectType,
xmin * scale_x,
ymin * scale_y,
xmax * scale_x,
ymax * scale_y,
prob,
difficult);
}
};
class ImageDescription {
public:
const std::list<DetectedObject> alist;
const bool check_probs;
explicit ImageDescription(const std::list<DetectedObject>& _alist, bool _check_probs = false)
: alist(_alist),
check_probs(_check_probs) {}
static float ioUMultiple(const ImageDescription& detectedObjects, const ImageDescription& desiredObjects) {
const ImageDescription *detectedObjectsSmall, *detectedObjectsBig;
bool check_probs = desiredObjects.check_probs;
if (detectedObjects.alist.size() < desiredObjects.alist.size()) {
detectedObjectsSmall = &detectedObjects;
detectedObjectsBig = &desiredObjects;
} else {
detectedObjectsSmall = &desiredObjects;
detectedObjectsBig = &detectedObjects;
}
std::list<DetectedObject> doS = detectedObjectsSmall->alist;
std::list<DetectedObject> doB = detectedObjectsBig->alist;
float fullScore = 0.0f;
while (doS.size() > 0) {
float score = 0.0f;
std::list<DetectedObject>::iterator bestJ = doB.end();
for (auto j = doB.begin(); j != doB.end(); j++) {
float curscore = DetectedObject::ioU(*doS.begin(), *j);
if (score < curscore) {
score = curscore;
bestJ = j;
}
}
float coeff = 1.0;
if (check_probs) {
if (bestJ != doB.end()) {
float mn = std::min((*bestJ).prob, (*doS.begin()).prob);
float mx = std::max((*bestJ).prob, (*doS.begin()).prob);
coeff = mn / mx;
}
}
doS.pop_front();
if (bestJ != doB.end())
doB.erase(bestJ);
fullScore += coeff * score;
}
fullScore /= detectedObjectsBig->alist.size();
return fullScore;
}
ImageDescription scale(float scale_x, float scale_y) const {
std::list<DetectedObject> slist;
for (auto& dob : alist) {
slist.push_back(dob.scale(scale_x, scale_y));
}
return ImageDescription(slist, check_probs);
}
};
struct AveragePrecisionCalculator {
private:
enum MatchKind { TruePositive, FalsePositive };
/**
* Here we count all TP and FP matches for all the classes in all the images.
*/
std::map<int, std::vector<std::pair<double, MatchKind>>> matches;
std::map<int, int> N;
double threshold;
static bool SortBBoxDescend(const DetectedObject& bbox1, const DetectedObject& bbox2) {
return bbox1.prob > bbox2.prob;
}
static bool SortPairDescend(const std::pair<double, MatchKind>& p1, const std::pair<double, MatchKind>& p2) {
return p1.first > p2.first;
}
public:
explicit AveragePrecisionCalculator(double _threshold) : threshold(_threshold) {}
// gt_bboxes -> des
// bboxes -> det
void consumeImage(const ImageDescription& detectedObjects, const ImageDescription& desiredObjects) {
// Collecting IoU values
std::vector<bool> visited(desiredObjects.alist.size(), false);
std::vector<DetectedObject> bboxes{std::begin(detectedObjects.alist), std::end(detectedObjects.alist)};
std::sort(bboxes.begin(), bboxes.end(), SortBBoxDescend);
for (auto&& detObj : bboxes) {
// Searching for the best match to this detection
// Searching for desired object
float overlap_max = -1;
int jmax = -1;
auto desmax = desiredObjects.alist.end();
int j = 0;
for (auto desObj = desiredObjects.alist.begin(); desObj != desiredObjects.alist.end(); desObj++, j++) {
double iou = DetectedObject::ioU(detObj, *desObj);
if (iou > overlap_max) {
overlap_max = static_cast<float>(iou);
jmax = j;
desmax = desObj;
}
}
MatchKind mk;
if (overlap_max >= threshold) {
if (!desmax->difficult) {
if (!visited[jmax]) {
mk = TruePositive;
visited[jmax] = true;
} else {
mk = FalsePositive;
}
matches[detObj.objectType].push_back(std::make_pair(detObj.prob, mk));
}
} else {
mk = FalsePositive;
matches[detObj.objectType].push_back(std::make_pair(detObj.prob, mk));
}
}
for (auto desObj = desiredObjects.alist.begin(); desObj != desiredObjects.alist.end(); desObj++) {
if (!desObj->difficult) {
N[desObj->objectType]++;
}
}
}
std::map<int, double> calculateAveragePrecisionPerClass() const {
/**
* Precision-to-TP curve per class (a variation of precision-to-recall curve without
* dividing into N)
*/
std::map<int, std::map<int, double>> precisionToTP;
std::map<int, double> res;
for (auto m : matches) {
// Sorting
std::sort(m.second.begin(), m.second.end(), SortPairDescend);
int clazz = m.first;
int TP = 0, FP = 0;
std::vector<double> prec;
std::vector<double> rec;
for (auto mm : m.second) {
// Here we are descending in a probability value
MatchKind mk = mm.second;
if (mk == TruePositive)
TP++;
else if (mk == FalsePositive)
FP++;
double precision = static_cast<double>(TP) / (TP + FP);
double recall = 0;
if (N.find(clazz) != N.end()) {
recall = static_cast<double>(TP) / N.at(clazz);
}
prec.push_back(precision);
rec.push_back(recall);
}
int num = static_cast<int>(rec.size());
// 11point from Caffe
double ap = 0;
std::vector<float> max_precs(11, 0.);
int start_idx = num - 1;
for (int j = 10; j >= 0; --j) {
for (int i = start_idx; i >= 0; --i) {
if (rec[i] < j / 10.) {
start_idx = i;
if (j > 0) {
max_precs[j - 1] = max_precs[j];
}
break;
} else {
if (max_precs[j] < prec[i]) {
max_precs[j] = static_cast<float>(prec[i]);
}
}
}
}
for (int j = 10; j >= 0; --j) {
ap += max_precs[j] / 11;
}
res[clazz] = ap;
}
return res;
}
};
/**
* @brief Adds colored rectangles to the image
* @param data - data where rectangles are put
* @param height - height of the rectangle
* @param width - width of the rectangle
* @param detectedObjects - vector of detected objects
*/
static UNUSED void addRectangles(unsigned char* data,
size_t height,
size_t width,
std::vector<DetectedObject> detectedObjects) {
std::vector<Color> colors = {{128, 64, 128}, {232, 35, 244}, {70, 70, 70}, {156, 102, 102}, {153, 153, 190},
{153, 153, 153}, {30, 170, 250}, {0, 220, 220}, {35, 142, 107}, {152, 251, 152},
{180, 130, 70}, {60, 20, 220}, {0, 0, 255}, {142, 0, 0}, {70, 0, 0},
{100, 60, 0}, {90, 0, 0}, {230, 0, 0}, {32, 11, 119}, {0, 74, 111},
{81, 0, 81}};
for (size_t i = 0; i < detectedObjects.size(); i++) {
int cls = detectedObjects[i].objectType % colors.size();
int xmin = static_cast<int>(detectedObjects[i].xmin * width);
int xmax = static_cast<int>(detectedObjects[i].xmax * width);
int ymin = static_cast<int>(detectedObjects[i].ymin * height);
int ymax = static_cast<int>(detectedObjects[i].ymax * height);
size_t shift_first = ymin * width * 3;
size_t shift_second = ymax * width * 3;
for (int x = xmin; x < xmax; x++) {
data[shift_first + x * 3] = colors.at(cls).red();
data[shift_first + x * 3 + 1] = colors.at(cls).green();
data[shift_first + x * 3 + 2] = colors.at(cls).blue();
data[shift_second + x * 3] = colors.at(cls).red();
data[shift_second + x * 3 + 1] = colors.at(cls).green();
data[shift_second + x * 3 + 2] = colors.at(cls).blue();
}
shift_first = xmin * 3;
shift_second = xmax * 3;
for (int y = ymin; y < ymax; y++) {
data[shift_first + y * width * 3] = colors.at(cls).red();
data[shift_first + y * width * 3 + 1] = colors.at(cls).green();
data[shift_first + y * width * 3 + 2] = colors.at(cls).blue();
data[shift_second + y * width * 3] = colors.at(cls).red();
data[shift_second + y * width * 3 + 1] = colors.at(cls).green();
data[shift_second + y * width * 3 + 2] = colors.at(cls).blue();
}
}
}
inline void showAvailableDevices() {
ov::Core core;
std::vector<std::string> devices = core.get_available_devices();
@@ -980,16 +352,6 @@ inline void showAvailableDevices() {
std::cout << std::endl;
}
/**
* @brief Parse text config file. The file must have the following format (with space a delimeter):
* CONFIG_NAME1 CONFIG_VALUE1
* CONFIG_NAME2 CONFIG_VALUE2
*
* @param configName - filename for a file with config options
* @param comment - lines starting with symbol `comment` are skipped
*/
std::map<std::string, std::string> parseConfig(const std::string& configName, char comment = '#');
inline std::string getFullDeviceName(ov::Core& core, std::string device) {
try {
return core.get_property(device, ov::device::full_name);
@@ -1066,15 +428,6 @@ static UNUSED void printPerformanceCounts(std::vector<ov::ProfilingInfo> perform
stream.flags(fmt);
}
static UNUSED void printPerformanceCounts(ov::InferRequest request,
std::ostream& stream,
std::string deviceName,
bool bshowHeader = true,
int precision = 3) {
auto performanceMap = request.get_profiling_info();
printPerformanceCounts(performanceMap, stream, deviceName, bshowHeader, precision);
}
static inline std::string double_to_string(const double number) {
std::stringstream ss;
ss << std::fixed << std::setprecision(2) << number;

92
samples/cpp/common/utils/include/samples/ocv_common.hpp
View File

@@ -1,92 +0,0 @@
// Copyright (C) 2018-2023 Intel Corporation
// SPDX-License-Identifier: Apache-2.0
//
/**
* @brief a header file with common samples functionality using OpenCV
* @file ocv_common.hpp
*/
#pragma once
#include <opencv2/opencv.hpp>
#include "openvino/openvino.hpp"
#include "samples/common.hpp"
/**
* @brief Sets image data stored in cv::Mat object to a given Blob object.
* @param orig_image - given cv::Mat object with an image data.
* @param blob - Blob object which to be filled by an image data.
* @param batchIndex - batch index of an image inside of the blob.
*/
template <typename T>
void matU8ToBlob(const cv::Mat& orig_image, InferenceEngine::Blob::Ptr& blob, int batchIndex = 0) {
InferenceEngine::SizeVector blobSize = blob->getTensorDesc().getDims();
const size_t width = blobSize[3];
const size_t height = blobSize[2];
const size_t channels = blobSize[1];
InferenceEngine::MemoryBlob::Ptr mblob = InferenceEngine::as<InferenceEngine::MemoryBlob>(blob);
OPENVINO_ASSERT(mblob,
"We expect blob to be inherited from MemoryBlob in matU8ToBlob, "
"but by fact we were not able to cast inputBlob to MemoryBlob");
// locked memory holder should be alive all time while access to its buffer happens
auto mblobHolder = mblob->wmap();
T* blob_data = mblobHolder.as<T*>();
cv::Mat resized_image(orig_image);
if (static_cast<int>(width) != orig_image.size().width || static_cast<int>(height) != orig_image.size().height) {
cv::resize(orig_image, resized_image, cv::Size(width, height));
}
int batchOffset = batchIndex * width * height * channels;
for (size_t c = 0; c < channels; c++) {
for (size_t h = 0; h < height; h++) {
for (size_t w = 0; w < width; w++) {
blob_data[batchOffset + c * width * height + h * width + w] = resized_image.at<cv::Vec3b>(h, w)[c];
}
}
}
}
/**
* @brief Wraps data stored inside of a passed cv::Mat object by new Blob pointer.
* @note: No memory allocation is happened. The blob just points to already existing
* cv::Mat data.
* @param mat - given cv::Mat object with an image data.
* @return resulting Blob pointer.
*/
static UNUSED InferenceEngine::Blob::Ptr wrapMat2Blob(const cv::Mat& mat) {
size_t channels = mat.channels();
size_t height = mat.size().height;
size_t width = mat.size().width;
size_t strideH = mat.step.buf[0];
size_t strideW = mat.step.buf[1];
bool is_dense = strideW == channels && strideH == channels * width;
OPENVINO_ASSERT(is_dense, "Doesn't support conversion from not dense cv::Mat");
InferenceEngine::TensorDesc tDesc(InferenceEngine::Precision::U8,
{1, channels, height, width},
InferenceEngine::Layout::NHWC);
return InferenceEngine::make_shared_blob<uint8_t>(tDesc, mat.data);
}
static UNUSED ov::Tensor wrapMat2Tensor(const cv::Mat& mat) {
const size_t channels = mat.channels();
const size_t height = mat.size().height;
const size_t width = mat.size().width;
const size_t strideH = mat.step.buf[0];
const size_t strideW = mat.step.buf[1];
const bool is_dense = strideW == channels && strideH == channels * width;
OPENVINO_ASSERT(is_dense, "Doesn't support conversion from not dense cv::Mat");
return ov::Tensor(ov::element::u8, ov::Shape{1, height, width, channels}, mat.data);
}

28
samples/cpp/common/utils/include/samples/vpu/vpu_tools_common.hpp
View File

@@ -1,28 +0,0 @@
// Copyright (C) 2018-2023 Intel Corporation
// SPDX-License-Identifier: Apache-2.0
//
#pragma once
#include <fstream>
#include <map>
#include <string>
static std::map<std::string, std::string> parseConfig(const std::string& configName, char comment = '#') {
std::map<std::string, std::string> config = {};
std::ifstream file(configName);
if (!file.is_open()) {
return config;
}
std::string key, value;
while (file >> key >> value) {
if (key.empty() || key[0] == comment) {
continue;
}
config[key] = value;
}
return config;
}

24
samples/cpp/common/utils/src/common.cpp
View File

@@ -1,24 +0,0 @@
// Copyright (C) 2018-2023 Intel Corporation
// SPDX-License-Identifier: Apache-2.0
//
#include "samples/common.hpp"
std::map<std::string, std::string> parseConfig(const std::string& configName, char comment) {
std::map<std::string, std::string> config = {};
std::ifstream file(configName);
if (!file.is_open()) {
return config;
}
std::string key, value;
while (file >> key >> value) {
if (key.empty() || key[0] == comment) {
continue;
}
config[key] = value;
}
return config;
}