Merge pull request #224 from saeugetier/220-reduced-resolution-neural-network-for-preview

220 reduced resolution neural network for preview

Author: saeugetier

qml/SettingsMenuForm.ui.qml

@@ -242,8 +242,8 @@ Item {
                             id: comboBoxNeuralNetworkRuntime
                             textRole: "text"
                             valueRole: "value"
-                            model: [{text: "ONNX Runtime", value: "ONNX"}, {text: "NCNN Runtime", value: "NCNN"}]
-                            Layout.preferredWidth: 200
+                            model: [{text: "ONNX Runtime", value: "ONNX"}, {text: "NCNN Runtime", value: "NCNN"}, {text: "NCNN Runtime (faster preview)", value: "NCNN_LOW_RES"} ]
+                            Layout.preferredWidth: 250
                         }
                     }
                 }

src/replacebackgroundvideofilter.cpp

@@ -57,6 +57,10 @@ void ReplaceBackgroundVideoFilter::setNeuralNetworkRuntime(QString runtime)
     {
         newRuntime = NeuralNetworkRuntime::ONNX;
     }
+    else if (runtime.contains("NCNN_LOW_RES"))
+    {
+        newRuntime = NeuralNetworkRuntime::NCNN_LOW_RES;
+    }
     else if (runtime.contains("NCNN"))
     {
         newRuntime = NeuralNetworkRuntime::NCNN;
@@ -111,6 +115,10 @@ QString ReplaceBackgroundVideoFilter::getNeuralNetworkRuntime() const
     {
         return QString("NCNN");
     }
+    else if (mNeuralNetworkRuntime == NeuralNetworkRuntime::NCNN_LOW_RES)
+    {
+        return QString("NCNN_LOW_RES");
+    }
     else
     {
         return QString("Unknown");
@@ -317,6 +325,12 @@ void ReplaceBackgroundFilterRunable::changeNeuralNetworkRuntime(const NeuralNetw
         mYoloSegmentorPreview.reset(new YOLOv11SegDetectorNcnn("yolo11n-seg_ncnn_model", "coco.names", false));
         mYoloSegmentorHighRes.reset(new YOLOv11SegDetectorNcnn("yolo11x-seg_ncnn_model", "coco.names", false));
     }
+    else if (runtime == NeuralNetworkRuntime::NCNN_LOW_RES)
+    {
+        qDebug() << "[INFO] Change YOLOv11Segmentation runtime to NCNN_LOW_RES";
+        mYoloSegmentorPreview.reset(new YOLOv11SegDetectorNcnn("yolo11n-seg_ncnn_model_320", "coco.names", false, true));
+        mYoloSegmentorHighRes.reset(new YOLOv11SegDetectorNcnn("yolo11x-seg_ncnn_model", "coco.names", false));
+    }
 }
 
 void ReplaceBackgroundFilterRunable::prepareBackground(cv::Mat &bg, cv::Size size)

src/replacebackgroundvideofilter.h

@@ -17,7 +17,8 @@ class ReplaceBackgroundFilterRunable;
 enum class NeuralNetworkRuntime
 {
     ONNX,
-    NCNN
+    NCNN,
+    NCNN_LOW_RES
 };
 
 class ReplaceBackgroundVideoFilter : public QVideoFrameInput

src/yolo11segncnn.cpp

@@ -9,7 +9,7 @@
 
 YOLOv11SegDetectorNcnn::YOLOv11SegDetectorNcnn(const std::string &modelPath,
                                                const std::string &labelsPath,
-                                               bool useGPU) : Yolo11Segementation(labelsPath)
+                                               bool useGPU, bool use320x320input) : Yolo11Segementation(labelsPath)
 {
     QString ressourcePathGeneric = QStandardPaths::locate(QStandardPaths::GenericDataLocation, "models", QStandardPaths::LocateDirectory);
     QString ressourcePathApp = QStandardPaths::locate(QStandardPaths::AppDataLocation, "models", QStandardPaths::LocateDirectory);
@@ -48,7 +48,14 @@ YOLOv11SegDetectorNcnn::YOLOv11SegDetectorNcnn(const std::string &modelPath,
     numOutputNodes = net.output_names().size();
 
     isDynamicInputShape = false;          // Assume static input shape by default. NCNN models typically have fixed input shapes.
-    inputImageShape = cv::Size(640, 640); // Default shape. This is fixed for YOLOv11SegNCNN
+    
+    if(use320x320input) {
+        inputImageShape = cv::Size(320, 320);
+    }
+    else
+    {
+        inputImageShape = cv::Size(640, 640); // Default shape. This is fixed for YOLOv11SegNCNN
+    }
 
     // Input
     if (numInputNodes != 1)
@@ -277,7 +284,7 @@ std::vector<Segmentation> YOLOv11SegDetectorNcnn::segment(const cv::Mat &image,
                      cv::Scalar(114, 114, 114), /*auto_=*/false,
                      /*scaleFill=*/false, /*scaleUp=*/true, /*stride=*/32);
 
-    ncnn::Mat in = ncnn::Mat::from_pixels_resize(letterboxImage.data, ncnn::Mat::PIXEL_BGR2RGB, letterboxImage.cols, letterboxImage.rows, 640, 640);
+    ncnn::Mat in = ncnn::Mat::from_pixels_resize(letterboxImage.data, ncnn::Mat::PIXEL_BGR2RGB, letterboxImage.cols, letterboxImage.rows, inputImageShape.width, inputImageShape.height);
 
     const float norm_vals[3] = {1 / 255.f, 1 / 255.f, 1 / 255.f};
     in.substract_mean_normalize(0, norm_vals);

src/yolo11segncnn.h

@@ -30,7 +30,7 @@ class YOLOv11SegDetectorNcnn : public Yolo11Segementation {
 public:
     YOLOv11SegDetectorNcnn(const std::string &modelPath,
                        const std::string &labelsPath,
-                       bool useGPU = false);
+                       bool useGPU = false, bool use320x320input = false);
 
     // Main API
     std::vector<Segmentation> segment(const cv::Mat &image,