Files
openpose/examples/tutorial_wrapper/2_user_synchronous.cpp
T
2017-06-16 19:35:37 -04:00

467 lines
26 KiB
C++

// ------------------------- OpenPose Library Tutorial - Thread - Example 2 - Synchronous -------------------------
// Synchronous mode: ideal for performance. The user can add his own frames producer / post-processor / consumer to the OpenPose wrapper or use the default ones.
// This example shows the user how to use the OpenPose wrapper class:
// 1. Extract and render keypoint / heatmap / PAF of that image
// 2. Save the results on disc
// 3. Display the rendered pose
// Everything in a multi-thread scenario
// In addition to the previous OpenPose modules, we also need to use:
// 1. `core` module:
// For the Array<float> class that the `pose` module needs
// For the Datum struct that the `thread` module sends between the queues
// 2. `utilities` module: for the error & logging functions, i.e. op::error & op::log respectively
// This file should only be used for the user to take specific examples.
// C++ std library dependencies
#include <atomic>
#include <chrono> // `std::chrono::` functions and classes, e.g. std::chrono::milliseconds
#include <cstdio> // sscanf
#include <string>
#include <thread> // std::this_thread
#include <vector>
// Other 3rdpary depencencies
#include <gflags/gflags.h> // DEFINE_bool, DEFINE_int32, DEFINE_int64, DEFINE_uint64, DEFINE_double, DEFINE_string
#include <glog/logging.h> // google::InitGoogleLogging
// OpenPose dependencies
// Option a) Importing all modules
#include <openpose/headers.hpp>
// Option b) Manually importing the desired modules. Recommended if you only intend to use a few modules.
// #include <openpose/core/headers.hpp>
// #include <openpose/experimental/headers.hpp>
// #include <openpose/face/headers.hpp>
// #include <openpose/filestream/headers.hpp>
// #include <openpose/gui/headers.hpp>
// #include <openpose/pose/headers.hpp>
// #include <openpose/producer/headers.hpp>
// #include <openpose/thread/headers.hpp>
// #include <openpose/utilities/headers.hpp>
// #include <openpose/wrapper/headers.hpp>
// See all the available parameter options withe the `--help` flag. E.g. `./build/examples/openpose/openpose.bin --help`.
// Note: This command will show you flags for other unnecessary 3rdparty files. Check only the flags for the OpenPose
// executable. E.g. for `openpose.bin`, look for `Flags from examples/openpose/openpose.cpp:`.
// Debugging
DEFINE_int32(logging_level, 3, "The logging level. Integer in the range [0, 255]. 0 will output any log() message, while"
" 255 will not output any. Current OpenPose library messages are in the range 0-4: 1 for"
" low priority messages and 4 for important ones.");
// Producer
DEFINE_string(image_dir, "examples/media/", "Process a directory of images.");
// OpenPose
DEFINE_string(model_folder, "models/", "Folder where the pose models (COCO and MPI) are located.");
DEFINE_string(resolution, "1280x720", "The image resolution (display and output). Use \"-1x-1\" to force the program to use the"
" default images resolution.");
DEFINE_int32(num_gpu, -1, "The number of GPU devices to use. If negative, it will use all the available GPUs in your"
" machine.");
DEFINE_int32(num_gpu_start, 0, "GPU device start number.");
DEFINE_int32(keypoint_scale, 0, "Scaling of the (x,y) coordinates of the final pose data array, i.e. the scale of the (x,y)"
" coordinates that will be saved with the `write_keypoint` & `write_keypoint_json` flags."
" Select `0` to scale it to the original source resolution, `1`to scale it to the net output"
" size (set with `net_resolution`), `2` to scale it to the final output size (set with"
" `resolution`), `3` to scale it in the range [0,1], and 4 for range [-1,1]. Non related"
" with `num_scales` and `scale_gap`.");
// OpenPose Body Pose
DEFINE_string(model_pose, "COCO", "Model to be used (e.g. COCO, MPI, MPI_4_layers).");
DEFINE_string(net_resolution, "656x368", "Multiples of 16. If it is increased, the accuracy usually increases. If it is decreased,"
" the speed increases.");
DEFINE_int32(num_scales, 1, "Number of scales to average.");
DEFINE_double(scale_gap, 0.3, "Scale gap between scales. No effect unless num_scales>1. Initial scale is always 1. If you"
" want to change the initial scale, you actually want to multiply the `net_resolution` by"
" your desired initial scale.");
DEFINE_bool(heatmaps_add_parts, false, "If true, it will add the body part heatmaps to the final op::Datum::poseHeatMaps array"
" (program speed will decrease). Not required for our library, enable it only if you intend"
" to process this information later. If more than one `add_heatmaps_X` flag is enabled, it"
" will place then in sequential memory order: body parts + bkg + PAFs. It will follow the"
" order on POSE_BODY_PART_MAPPING in `include/openpose/pose/poseParameters.hpp`.");
DEFINE_bool(heatmaps_add_bkg, false, "Same functionality as `add_heatmaps_parts`, but adding the heatmap corresponding to"
" background.");
DEFINE_bool(heatmaps_add_PAFs, false, "Same functionality as `add_heatmaps_parts`, but adding the PAFs.");
DEFINE_int32(heatmaps_scale, 2, "Set 0 to scale op::Datum::poseHeatMaps in the range [0,1], 1 for [-1,1]; and 2 for integer"
" rounded [0,255].");
// OpenPose Face
DEFINE_bool(face, false, "Enables face keypoint detection. It will share some parameters from the body pose, e.g."
" `model_folder`.");
DEFINE_string(face_net_resolution, "368x368", "Multiples of 16. Analogous to `net_resolution` but applied to the face keypoint detector."
" 320x320 usually works fine while giving a substantial speed up when multiple faces on the"
" image.");
// OpenPose Hand
// OpenPose Rendering
DEFINE_int32(part_to_show, 0, "Part to show from the start.");
DEFINE_bool(disable_blending, false, "If blending is enabled, it will merge the results with the original frame. If disabled, it"
" will only display the results.");
// OpenPose Rendering Pose
DEFINE_int32(render_pose, 1, "Set to 0 for no rendering, 1 for CPU rendering (slightly faster), and 2 for GPU rendering"
" (slower but greater functionality, e.g. `alpha_X` flags). If rendering is enabled, it will"
" render both `outputData` and `cvOutputData` with the original image and desired body part"
" to be shown (i.e. keypoints, heat maps or PAFs).");
DEFINE_double(alpha_pose, 0.6, "Blending factor (range 0-1) for the body part rendering. 1 will show it completely, 0 will"
" hide it. Only valid for GPU rendering.");
DEFINE_double(alpha_heatmap, 0.7, "Blending factor (range 0-1) between heatmap and original frame. 1 will only show the"
" heatmap, 0 will only show the frame. Only valid for GPU rendering.");
// OpenPose Rendering Face
DEFINE_int32(render_face, -1, "Analogous to `render_pose` but applied to the face. Extra option: -1 to use the same"
" configuration that `render_pose` is using.");
DEFINE_double(alpha_face, 0.6, "Analogous to `alpha_pose` but applied to face.");
DEFINE_double(alpha_heatmap_face, 0.7, "Analogous to `alpha_heatmap` but applied to face.");
// Result Saving
DEFINE_string(write_images, "", "Directory to write rendered frames in `write_images_format` image format.");
DEFINE_string(write_images_format, "png", "File extension and format for `write_images`, e.g. png, jpg or bmp. Check the OpenCV"
" function cv::imwrite for all compatible extensions.");
DEFINE_string(write_video, "", "Full file path to write rendered frames in motion JPEG video format. It might fail if the"
" final path does not finish in `.avi`. It internally uses cv::VideoWriter.");
DEFINE_string(write_keypoint, "", "Directory to write the people pose keypoint data. Format with `write_keypoint_format`.");
DEFINE_string(write_keypoint_format, "yml", "File extension and format for `write_keypoint`: json, xml, yaml & yml. Json not available"
" for OpenCV < 3.0, use `write_keypoint_json` instead.");
DEFINE_string(write_keypoint_json, "", "Directory to write people pose data in *.json format, compatible with any OpenCV version.");
DEFINE_string(write_coco_json, "", "Full file path to write people pose data with *.json COCO validation format.");
DEFINE_string(write_heatmaps, "", "Directory to write heatmaps in *.png format. At least 1 `add_heatmaps_X` flag must be"
" enabled.");
DEFINE_string(write_heatmaps_format, "png", "File extension and format for `write_heatmaps`, analogous to `write_images_format`."
" Recommended `png` or any compressed and lossless format.");
// If the user needs his own variables, he can inherit the op::Datum struct and add them
// UserDatum can be directly used by the OpenPose wrapper because it inherits from op::Datum, just define Wrapper<UserDatum> instead of
// Wrapper<op::Datum>
struct UserDatum : public op::Datum
{
bool boolThatUserNeedsForSomeReason;
UserDatum(const bool boolThatUserNeedsForSomeReason_ = false) :
boolThatUserNeedsForSomeReason{boolThatUserNeedsForSomeReason_}
{}
};
// The W-classes can be implemented either as a template or as simple classes given
// that the user usually knows which kind of data he will move between the queues,
// in this case we assume a std::shared_ptr of a std::vector of UserDatum
// This worker will just read and return all the jpg files in a directory
class WUserInput : public op::WorkerProducer<std::shared_ptr<std::vector<UserDatum>>>
{
public:
WUserInput(const std::string& directoryPath) :
mImageFiles{op::getFilesOnDirectory(directoryPath, "jpg")},
// mImageFiles{op::getFilesOnDirectory(directoryPath, std::vector<std::string>{"jpg", "png"})}, // If we want "jpg" + "png" images
mCounter{0}
{
if (mImageFiles.empty())
op::error("No images found on: " + directoryPath, __LINE__, __FUNCTION__, __FILE__);
}
void initializationOnThread() {}
std::shared_ptr<std::vector<UserDatum>> workProducer()
{
try
{
// Close program when empty frame
if (mImageFiles.size() <= mCounter)
{
op::log("Last frame read and added to queue. Closing program after it is processed.", op::Priority::Max);
// This funtion stops this worker, which will eventually stop the whole thread system once all the frames have been processed
this->stop();
return nullptr;
}
else
{
// Create new datum
auto datumsPtr = std::make_shared<std::vector<UserDatum>>();
datumsPtr->emplace_back();
auto& datum = datumsPtr->at(0);
// Fill datum
datum.cvInputData = cv::imread(mImageFiles.at(mCounter++));
// If empty frame -> return nullptr
if (datum.cvInputData.empty())
{
op::log("Empty frame detected on path: " + mImageFiles.at(mCounter-1) + ". Closing program.", op::Priority::Max);
this->stop();
datumsPtr = nullptr;
}
return datumsPtr;
}
}
catch (const std::exception& e)
{
op::log("Some kind of unexpected error happened.");
this->stop();
op::error(e.what(), __LINE__, __FUNCTION__, __FILE__);
return nullptr;
}
}
private:
const std::vector<std::string> mImageFiles;
unsigned long long mCounter;
};
// This worker will just invert the image
class WUserPostProcessing : public op::Worker<std::shared_ptr<std::vector<UserDatum>>>
{
public:
WUserPostProcessing()
{
// User's constructor here
}
void initializationOnThread() {}
void work(std::shared_ptr<std::vector<UserDatum>>& datumsPtr)
{
// User's post-processing (after OpenPose processing & before OpenPose outputs) here
// datum.cvOutputData: rendered frame with pose or heatmaps
// datum.poseKeypoints: Array<float> with the estimated pose
try
{
if (datumsPtr != nullptr && !datumsPtr->empty())
for (auto& datum : *datumsPtr)
cv::bitwise_not(datum.cvOutputData, datum.cvOutputData);
}
catch (const std::exception& e)
{
op::log("Some kind of unexpected error happened.");
this->stop();
op::error(e.what(), __LINE__, __FUNCTION__, __FILE__);
}
}
};
// This worker will just read and return all the jpg files in a directory
class WUserOutput : public op::WorkerConsumer<std::shared_ptr<std::vector<UserDatum>>>
{
public:
void initializationOnThread() {}
void workConsumer(const std::shared_ptr<std::vector<UserDatum>>& datumsPtr)
{
try
{
// User's displaying/saving/other processing here
// datum.cvOutputData: rendered frame with pose or heatmaps
// datum.poseKeypoints: Array<float> with the estimated pose
if (datumsPtr != nullptr && !datumsPtr->empty())
{
cv::imshow("User worker GUI", datumsPtr->at(0).cvOutputData);
cv::waitKey(1); // It displays the image and sleeps at least 1 ms (it usually sleeps ~5-10 msec to display the image)
}
}
catch (const std::exception& e)
{
op::log("Some kind of unexpected error happened.");
this->stop();
op::error(e.what(), __LINE__, __FUNCTION__, __FILE__);
}
}
};
op::PoseModel gflagToPoseModel(const std::string& poseModeString)
{
op::log("", op::Priority::Low, __LINE__, __FUNCTION__, __FILE__);
if (poseModeString == "COCO")
return op::PoseModel::COCO_18;
else if (poseModeString == "MPI")
return op::PoseModel::MPI_15;
else if (poseModeString == "MPI_4_layers")
return op::PoseModel::MPI_15_4;
else
{
op::error("String does not correspond to any model (COCO, MPI, MPI_4_layers)", __LINE__, __FUNCTION__, __FILE__);
return op::PoseModel::COCO_18;
}
}
op::ScaleMode gflagToScaleMode(const int keypointScale)
{
op::log("", op::Priority::Low, __LINE__, __FUNCTION__, __FILE__);
if (keypointScale == 0)
return op::ScaleMode::InputResolution;
else if (keypointScale == 1)
return op::ScaleMode::NetOutputResolution;
else if (keypointScale == 2)
return op::ScaleMode::OutputResolution;
else if (keypointScale == 3)
return op::ScaleMode::ZeroToOne;
else if (keypointScale == 4)
return op::ScaleMode::PlusMinusOne;
else
{
const std::string message = "String does not correspond to any scale mode: (0, 1, 2, 3, 4) for (InputResolution,"
" NetOutputResolution, OutputResolution, ZeroToOne, PlusMinusOne).";
op::error(message, __LINE__, __FUNCTION__, __FILE__);
return op::ScaleMode::InputResolution;
}
}
std::vector<op::HeatMapType> gflagToHeatMaps(const bool heatMapsAddParts, const bool heatMapsAddBkg, const bool heatMapsAddPAFs)
{
std::vector<op::HeatMapType> heatMapTypes;
if (heatMapsAddParts)
heatMapTypes.emplace_back(op::HeatMapType::Parts);
if (heatMapsAddBkg)
heatMapTypes.emplace_back(op::HeatMapType::Background);
if (heatMapsAddPAFs)
heatMapTypes.emplace_back(op::HeatMapType::PAFs);
return heatMapTypes;
}
op::RenderMode gflagToRenderMode(const int renderFlag, const int renderPoseFlag = -2)
{
if (renderFlag == -1 && renderPoseFlag != -2)
return gflagToRenderMode(renderPoseFlag, -2);
else if (renderFlag == 0)
return op::RenderMode::None;
else if (renderFlag == 1)
return op::RenderMode::Cpu;
else if (renderFlag == 2)
return op::RenderMode::Gpu;
else
{
op::error("Undefined RenderMode selected.", __LINE__, __FUNCTION__, __FILE__);
return op::RenderMode::None;
}
}
// Google flags into program variables
std::tuple<op::Point<int>, op::Point<int>, op::Point<int>, op::PoseModel, op::ScaleMode, std::vector<op::HeatMapType>,
op::ScaleMode> gflagsToOpParameters()
{
op::log("", op::Priority::Low, __LINE__, __FUNCTION__, __FILE__);
// outputSize
op::Point<int> outputSize;
auto nRead = sscanf(FLAGS_resolution.c_str(), "%dx%d", &outputSize.x, &outputSize.y);
op::checkE(nRead, 2, "Error, resolution format (" + FLAGS_resolution + ") invalid, should be e.g., 960x540 ",
__LINE__, __FUNCTION__, __FILE__);
// netInputSize
op::Point<int> netInputSize;
nRead = sscanf(FLAGS_net_resolution.c_str(), "%dx%d", &netInputSize.x, &netInputSize.y);
op::checkE(nRead, 2, "Error, net resolution format (" + FLAGS_net_resolution + ") invalid, should be e.g., 656x368 (multiples of 16)",
__LINE__, __FUNCTION__, __FILE__);
// faceNetInputSize
op::Point<int> faceNetInputSize;
nRead = sscanf(FLAGS_face_net_resolution.c_str(), "%dx%d", &faceNetInputSize.x, &faceNetInputSize.y);
op::checkE(nRead, 2, "Error, face net resolution format (" + FLAGS_face_net_resolution
+ ") invalid, should be e.g., 368x368 (multiples of 16)", __LINE__, __FUNCTION__, __FILE__);
// poseModel
const auto poseModel = gflagToPoseModel(FLAGS_model_pose);
// keypointScale
const auto keypointScale = gflagToScaleMode(FLAGS_keypoint_scale);
// heatmaps to add
const auto heatMapTypes = gflagToHeatMaps(FLAGS_heatmaps_add_parts, FLAGS_heatmaps_add_bkg, FLAGS_heatmaps_add_PAFs);
op::check(FLAGS_heatmaps_scale >= 0 && FLAGS_heatmaps_scale <= 2, "Non valid `heatmaps_scale`.", __LINE__, __FUNCTION__, __FILE__);
const auto heatMapScale = (FLAGS_heatmaps_scale == 0 ? op::ScaleMode::PlusMinusOne
: (FLAGS_heatmaps_scale == 1 ? op::ScaleMode::ZeroToOne : op::ScaleMode::UnsignedChar ));
// return
return std::make_tuple(outputSize, netInputSize, faceNetInputSize, poseModel, keypointScale, heatMapTypes, heatMapScale);
}
int openPoseTutorialWrapper2()
{
// logging_level
op::check(0 <= FLAGS_logging_level && FLAGS_logging_level <= 255, "Wrong logging_level value.", __LINE__, __FUNCTION__, __FILE__);
op::ConfigureLog::setPriorityThreshold((op::Priority)FLAGS_logging_level);
// op::ConfigureLog::setPriorityThreshold(op::Priority::None); // To print all logging messages
op::log("Starting pose estimation demo.", op::Priority::Max);
const auto timerBegin = std::chrono::high_resolution_clock::now();
// Applying user defined configuration
op::Point<int> outputSize;
op::Point<int> netInputSize;
op::Point<int> faceNetInputSize;
op::PoseModel poseModel;
op::ScaleMode keypointScale;
std::vector<op::HeatMapType> heatMapTypes;
op::ScaleMode heatMapScale;
std::tie(outputSize, netInputSize, faceNetInputSize, poseModel, keypointScale, heatMapTypes, heatMapScale) = gflagsToOpParameters();
op::log("", op::Priority::Low, __LINE__, __FUNCTION__, __FILE__);
// Initializing the user custom classes
// Frames producer (e.g. video, webcam, ...)
auto wUserInput = std::make_shared<WUserInput>(FLAGS_image_dir);
// Processing
auto wUserPostProcessing = std::make_shared<WUserPostProcessing>();
// GUI (Display)
auto wUserOutput = std::make_shared<WUserOutput>();
op::Wrapper<std::vector<UserDatum>> opWrapper;
// Add custom input
const auto workerInputOnNewThread = false;
opWrapper.setWorkerInput(wUserInput, workerInputOnNewThread);
// Add custom processing
const auto workerProcessingOnNewThread = false;
opWrapper.setWorkerPostProcessing(wUserPostProcessing, workerProcessingOnNewThread);
// Add custom output
const auto workerOutputOnNewThread = true;
opWrapper.setWorkerOutput(wUserOutput, workerOutputOnNewThread);
// Configure OpenPose
const bool displayGui = false;
const bool guiVerbose = false;
const bool fullScreen = false;
const op::WrapperStructPose wrapperStructPose{netInputSize, outputSize, keypointScale, FLAGS_num_gpu, FLAGS_num_gpu_start,
FLAGS_num_scales, (float)FLAGS_scale_gap, gflagToRenderMode(FLAGS_render_pose), poseModel,
!FLAGS_disable_blending, (float)FLAGS_alpha_pose, (float)FLAGS_alpha_heatmap,
FLAGS_part_to_show, FLAGS_model_folder, heatMapTypes, heatMapScale};
// Face configuration (use op::WrapperStructFace{} to disable it)
const op::WrapperStructFace wrapperStructFace{FLAGS_face, faceNetInputSize, gflagToRenderMode(FLAGS_render_face, FLAGS_render_pose),
(float)FLAGS_alpha_face, (float)FLAGS_alpha_heatmap_face};
// Hand configuration (use op::WrapperStructHand{} to disable it)
const op::experimental::WrapperStructHand wrapperStructHand{false};
// Consumer (comment or use default argument to disable any output)
const op::WrapperStructOutput wrapperStructOutput{displayGui, guiVerbose, fullScreen, FLAGS_write_keypoint,
op::stringToDataFormat(FLAGS_write_keypoint_format), FLAGS_write_keypoint_json,
FLAGS_write_coco_json, FLAGS_write_images, FLAGS_write_images_format, FLAGS_write_video,
FLAGS_write_heatmaps, FLAGS_write_heatmaps_format};
// Configure wrapper
opWrapper.configure(wrapperStructPose, wrapperStructFace, wrapperStructHand, op::WrapperStructInput{}, wrapperStructOutput);
// Set to single-thread running (e.g. for debugging purposes)
// opWrapper.disableMultiThreading();
op::log("Starting thread(s)", op::Priority::Max);
// Two different ways of running the program on multithread environment
// // Option a) Recommended - Also using the main thread (this thread) for processing (it saves 1 thread)
// // Start, run & stop threads
opWrapper.exec(); // It blocks this thread until all threads have finished
// Option b) Keeping this thread free in case you want to do something else meanwhile, e.g. profiling the GPU memory
// // VERY IMPORTANT NOTE: if OpenCV is compiled with Qt support, this option will not work. Qt needs the main thread to
// // plot visual results, so the final GUI (which uses OpenCV) would return an exception similar to:
// // `QMetaMethod::invoke: Unable to invoke methods with return values in queued connections`
// // Start threads
// opWrapper.start();
// // Profile used GPU memory
// // 1: wait ~10sec so the memory has been totally loaded on GPU
// // 2: profile the GPU memory
// std::this_thread::sleep_for(std::chrono::milliseconds{1000});
// op::log("Random task here...", op::Priority::Max);
// // Keep program alive while running threads
// while (opWrapper.isRunning())
// std::this_thread::sleep_for(std::chrono::milliseconds{33});
// // Stop and join threads
// op::log("Stopping thread(s)", op::Priority::Max);
// opWrapper.stop();
// Measuring total time
const auto now = std::chrono::high_resolution_clock::now();
const auto totalTimeSec = (double)std::chrono::duration_cast<std::chrono::nanoseconds>(now-timerBegin).count() * 1e-9;
const auto message = "Real-time pose estimation demo successfully finished. Total time: " + std::to_string(totalTimeSec) + " seconds.";
op::log(message, op::Priority::Max);
return 0;
}
int main(int argc, char *argv[])
{
// Initializing google logging (Caffe uses it for logging)
google::InitGoogleLogging("openPoseTutorialWrapper2");
// Parsing command line flags
gflags::ParseCommandLineFlags(&argc, &argv, true);
// Running openPoseTutorialWrapper2
return openPoseTutorialWrapper2();
}