75 lines
3.2 KiB
C++
75 lines
3.2 KiB
C++
#include "../nodes/vp_app_src_node.h"
|
||
#include "../nodes/infers/vp_yunet_face_detector_node.h"
|
||
#include "../nodes/osd/vp_face_osd_node.h"
|
||
#include "../nodes/vp_app_des_node.h"
|
||
|
||
#include "../utils/analysis_board/vp_analysis_board.h"
|
||
|
||
/*
|
||
* ## app_src_des_sample ##
|
||
* 1. receive images from host code， based on vp_app_src_node
|
||
* 2. detect faces and draw results
|
||
* 3. display on screen in host code again using cv::imshow(...) and print rectangles and keypoints of face, based on vp_app_des_node
|
||
* we treat VideoPipe(pipeline) as a simple face detector tool in this sample.
|
||
*/
|
||
|
||
int main() {
|
||
VP_SET_LOG_LEVEL(vp_utils::vp_log_level::INFO);
|
||
VP_LOGGER_INIT();
|
||
|
||
// create nodes
|
||
auto app_src_0 = std::make_shared<vp_nodes::vp_app_src_node>("app_src_0", 0);
|
||
auto yunet_face_detector_0 = std::make_shared<vp_nodes::vp_yunet_face_detector_node>("yunet_face_detector_0", "./vp_data/models/face/face_detection_yunet_2022mar.onnx");
|
||
auto osd_0 = std::make_shared<vp_nodes::vp_face_osd_node>("osd_0");
|
||
auto app_des_0 = std::make_shared<vp_nodes::vp_app_des_node>("app_des_0", 0);
|
||
|
||
// register callback to display result
|
||
std::string ori_win_title = "original frame using cv::imshow(...)";
|
||
std::string osd_win_title = "osd frame using cv::imshow(...)";
|
||
cv::namedWindow(ori_win_title,cv::WindowFlags::WINDOW_NORMAL);
|
||
cv::namedWindow(osd_win_title,cv::WindowFlags::WINDOW_NORMAL);
|
||
app_des_0->set_app_des_result_hooker([&](std::string node_name, std::shared_ptr<vp_objects::vp_meta> meta) {
|
||
// only deal with frame meta
|
||
if (meta->meta_type == vp_objects::vp_meta_type::FRAME) {
|
||
auto frame_meta = std::dynamic_pointer_cast<vp_objects::vp_frame_meta>(meta);
|
||
cv::imshow(ori_win_title, frame_meta->frame);
|
||
|
||
// osd frame may be empty
|
||
if (!frame_meta->osd_frame.empty()) {
|
||
cv::imshow(osd_win_title, frame_meta->osd_frame);
|
||
}
|
||
|
||
// print rectangle and keypoints of face
|
||
std::cout << "detected " << frame_meta->face_targets.size() << " faces:(print from host code)" << std::endl;
|
||
for (int i = 0; i < frame_meta->face_targets.size(); ++i) {
|
||
auto f = frame_meta->face_targets[i];
|
||
std::cout << i << ": [" << f->x << "," << f->y << "," << f->width << "," << f->height << "] [";
|
||
for (auto k: f->key_points) {
|
||
std::cout << "(" << k.first << "," << k.second << ")";
|
||
}
|
||
std::cout << "]" << std::endl;
|
||
}
|
||
}
|
||
});
|
||
|
||
// construct pipeline
|
||
yunet_face_detector_0->attach_to({app_src_0});
|
||
osd_0->attach_to({yunet_face_detector_0});
|
||
app_des_0->attach_to({osd_0});
|
||
|
||
// start pipeline
|
||
app_src_0->start();
|
||
|
||
// for debug purpose
|
||
vp_utils::vp_analysis_board board({app_src_0});
|
||
board.display(1, false); // no block
|
||
|
||
// read image in host code manually
|
||
auto img = cv::imread("./vp_data/test_images/faces/swap/2mans.jpg");
|
||
// push image to pipeline in host code manually
|
||
app_src_0->push_frames({img});
|
||
|
||
// pipeline will process image asynchronously and display result here...
|
||
std::string wait;
|
||
std:getline(std::cin, wait);
|
||
} |