add detect-camera example for onnx

example/opencv_dnn/README.md

@@ -33,5 +33,6 @@ You can view the network architecture here[[netron]](https://netron.app/?url=htt
     ```
 2. Run the example:
     ```shell
-    ./detect /path/to/image /path/to/YuFaceDetectNet.onnx
+    ./detect-image /path/to/image /path/to/YuFaceDetectNet.onnx
+    ./detect-camera 0 /path/to/YuFaceDetectNet.onnx  # '0' is your camera index.
     ```

example/opencv_dnn/cpp/CMakeLists.txt

@@ -6,5 +6,8 @@ project(libfacedetection_opencvdnn)
 find_package(OpenCV 4.5.1 REQUIRED)
 include_directories(${OpenCV_INCLUDE_DIRS})
 
-add_executable(detect detect.cpp priorbox.cpp utils.cpp)
-target_link_libraries(detect ${OpenCV_LIBS})
+add_executable(detect-image detect-image.cpp priorbox.cpp utils.cpp)
+target_link_libraries(detect-image ${OpenCV_LIBS})
+
+add_executable(detect-camera detect-camera.cpp priorbox.cpp utils.cpp)
+target_link_libraries(detect-camera ${OpenCV_LIBS})

example/opencv_dnn/cpp/detect-camera.cpp

@@ -0,0 +1,95 @@
+#include <vector>
+#include <string>
+#include <iostream>
+
+#include "priorbox.hpp"
+#include "utils.hpp"
+
+#include "opencv2/opencv.hpp"
+
+int main(int argc, char* argv[]) {
+
+    if(argc != 3)
+    {
+        printf("Usage: %s <camera index>\n", argv[0]);
+        return -1;
+    }
+
+    // Load .onnx model using OpenCV's DNN module
+    cv::dnn::Net net = cv::dnn::readNet(argv[2]);
+    net.setPreferableBackend(cv::dnn::DNN_BACKEND_DEFAULT);
+    net.setPreferableTarget(cv::dnn::DNN_TARGET_CPU);
+
+    cv::VideoCapture cap;
+    cv::Mat im, blob;
+    cv::TickMeter cvtm;
+    cv::Size img_shape;
+    std::vector<cv::String> output_names = { "loc", "conf", "iou" };
+    std::vector<cv::Mat> output_blobs;
+
+    cv::String title = cv::String("Detection Results on") + cv::String(argv[1]);
+
+    // Inference hyperparameters
+    float conf_thresh = 0.6;
+    float nms_thresh = 0.3;
+    int keep_top_k = 750;
+    // Result
+    bool vis = false;
+
+    if( isdigit(argv[1][0]))
+    {
+        cap.open(argv[1][0]-'0');
+        if(! cap.isOpened())
+        {
+            std::cerr << "Cannot open the camera." << std::endl;
+            return 0;
+        }
+    }
+
+    cap >> im;
+    img_shape = im.size();
+    PriorBox pb(img_shape, img_shape);
+    std::vector<Face> dets;
+
+    if( cap.isOpened())
+    {
+        while(true)
+        {
+            cap >> im;
+            cvtm.start();
+
+            // Build blob
+            blob = cv::dnn::blobFromImage(im, 1.0,  cv::Size(), cv::Scalar());
+
+            // Forward
+            net.setInput(blob);
+            net.forward(output_blobs, output_names);
+
+            // Decode bboxes, landmarks and scores
+            dets = pb.decode(output_blobs[0], output_blobs[1], output_blobs[2], conf_thresh);
+
+            // NMS
+            if (dets.size() > 1) {
+                nms(dets, nms_thresh);
+                if (dets.size() > keep_top_k) { dets.erase(dets.begin()+keep_top_k, dets.end()); }
+            }
+            else if (dets.size() < 1) {
+                std::cout << "No faces found." << std::endl;
+            }
+
+            cvtm.stop();
+
+            std::string timeLabel = cv::format("Inference time: %.2f ms", cvtm.getTimeMilli());
+            cv::putText(im, timeLabel, cv::Point(0, 15), cv::FONT_HERSHEY_SIMPLEX, 0.5, cv::Scalar(0, 255, 0), 1);
+
+            draw(im, dets);
+
+            cvtm.reset();
+            cv::imshow(title, im);
+            if((cv::waitKey(1)& 0xFF) == 27)
+                break;
+        }
+    }
+
+    return 0;
+}