feature: add video tracker implementation using Vit DNN

Signed-off-by: deadprogram <ron@hybridgroup.com>

.github/workflows/linux.yml

@@ -46,6 +46,7 @@ jobs:
           curl -sL https://github.com/onnx/models/raw/main/validated/vision/classification/inception_and_googlenet/googlenet/model/googlenet-9.onnx > ${GITHUB_WORKSPACE}/testdata/googlenet-9.onnx
           curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_recognition_sface/face_recognition_sface_2021dec.onnx > ${GITHUB_WORKSPACE}/testdata/face_recognition_sface_2021dec.onnx
           curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_detection_yunet/face_detection_yunet_2023mar.onnx > ${GITHUB_WORKSPACE}/testdata/face_detection_yunet_2023mar.onnx
+          curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/object_tracking_vittrack/object_tracking_vittrack_2023sep.onnx > ${GITHUB_WORKSPACE}/testdata/object_tracking_vittrack_2023sep.onnx
       - name: Run main tests
         run: xvfb-run -a --error-file /var/log/xvfb_error.log --server-args="-screen 0 1024x768x24 +extension RANDR" go test -v -coverprofile=/tmp/coverage.out -count=1 -tags matprofile .
         env:

.github/workflows/macos.yml

@@ -56,6 +56,7 @@ jobs:
           curl -sL https://github.com/onnx/models/raw/main/validated/vision/classification/inception_and_googlenet/googlenet/model/googlenet-9.onnx > ${GITHUB_WORKSPACE}/testdata/googlenet-9.onnx
           curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_recognition_sface/face_recognition_sface_2021dec.onnx > ${GITHUB_WORKSPACE}/testdata/face_recognition_sface_2021dec.onnx
           curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_detection_yunet/face_detection_yunet_2023mar.onnx > ${GITHUB_WORKSPACE}/testdata/face_detection_yunet_2023mar.onnx
+          curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/object_tracking_vittrack/object_tracking_vittrack_2023sep.onnx > ${GITHUB_WORKSPACE}/testdata/object_tracking_vittrack_2023sep.onnx
       - name: Run main tests
         run: go test -v -tags matprofile .
         env:

.github/workflows/windows.yml

@@ -92,6 +92,7 @@ jobs:
           curl -sL https://github.com/onnx/models/raw/main/validated/vision/classification/inception_and_googlenet/googlenet/model/googlenet-9.onnx > ./testdata/googlenet-9.onnx
           curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_recognition_sface/face_recognition_sface_2021dec.onnx > ./testdata/face_recognition_sface_2021dec.onnx
           curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_detection_yunet/face_detection_yunet_2023mar.onnx > ./testdata/face_detection_yunet_2023mar.onnx
+          curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/object_tracking_vittrack/object_tracking_vittrack_2023sep.onnx > ./testdata/object_tracking_vittrack_2023sep.onnx
       - name: Install GOTURN test model
         shell: bash
         run: |

Makefile

@@ -362,6 +362,7 @@ download_onnx_testdata: create_testdata_dir
 	curl -sL https://github.com/onnx/models/raw/main/validated/vision/classification/inception_and_googlenet/googlenet/model/googlenet-9.onnx > ./testdata/googlenet-9.onnx
 	curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_recognition_sface/face_recognition_sface_2021dec.onnx > ./testdata/face_recognition_sface_2021dec.onnx
 	curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/face_detection_yunet/face_detection_yunet_2023mar.onnx > ./testdata/face_detection_yunet_2023mar.onnx
+	curl -sL https://github.com/opencv/opencv_zoo/raw/refs/heads/main/models/object_tracking_vittrack/object_tracking_vittrack_2023sep.onnx > ./testdata/object_tracking_vittrack_2023sep.onnx
 
 download_goturn_testdata: create_testdata_dir
 	curl -sL https://raw.githubusercontent.com/opencv/opencv_extra/c4219d5eb3105ed8e634278fad312a1a8d2c182d/testdata/tracking/goturn.prototxt > ./testdata/goturn.prototxt

video.cpp

@@ -190,6 +190,31 @@ void TrackerGOTURN_Close(TrackerGOTURN tr) {
     delete tr;
 }
 
+TrackerVit TrackerVit_Create() {
+    try {
+		return new cv::Ptr<cv::TrackerVit>(cv::TrackerVit::create());
+    } catch(const cv::Exception& e){
+        setExceptionInfo(e.code, e.what());
+		return NULL;
+    }
+}
+
+TrackerVit TrackerVit_CreateWithParams(const char* model){
+    try {
+		cv::TrackerVit::Params params;
+		params.net = model;
+	  
+		return new cv::Ptr<cv::TrackerVit>(cv::TrackerVit::create(params));
+	} catch(const cv::Exception& e){
+        setExceptionInfo(e.code, e.what());
+		return NULL;
+    }
+}
+
+void TrackerVit_Close(TrackerVit self) {
+    delete self;
+}
+
 KalmanFilter KalmanFilter_New(int dynamParams, int measureParams) {
     try {
 		return new cv::KalmanFilter(dynamParams, measureParams, 0, CV_32F);

video.go

@@ -301,6 +301,43 @@ func (t TrackerGOTURN) Close() error {
 
 }
 
+// TrackerVit is a Tracker using the Vit tracker, a super lightweight dnn-based general object tracking.
+//
+// For further details, please see:
+// https://docs.opencv.org/4.x/d9/d26/classcv_1_1TrackerVit.html
+type TrackerVit struct {
+	p C.TrackerVit
+}
+
+// NewTrackerVit returns a new TrackerVit.
+func NewTrackerVit() Tracker {
+	return TrackerVit{p: C.TrackerVit_Create()}
+}
+
+func NewTrackerVitWithParams(model string) TrackerVit {
+	c_model := C.CString(model)
+	defer C.free(unsafe.Pointer(c_model))
+
+	return TrackerVit{p: C.TrackerVit_CreateWithParams(c_model)}
+}
+
+// Close closes the TrackerMIL.
+func (trk TrackerVit) Close() error {
+	C.TrackerVit_Close(trk.p)
+	trk.p = nil
+	return nil
+}
+
+// Init initializes the TrackerVIT.
+func (trk TrackerVit) Init(img Mat, boundingBox image.Rectangle) bool {
+	return trackerInit(C.Tracker(trk.p), img, boundingBox)
+}
+
+// Update updates the TrackerVIT.
+func (trk TrackerVit) Update(img Mat) (image.Rectangle, bool) {
+	return trackerUpdate(C.Tracker(trk.p), img)
+}
+
 // KalmanFilter implements a standard Kalman filter http://en.wikipedia.org/wiki/Kalman_filter.
 // However, you can modify transitionMatrix, controlMatrix, and measurementMatrix
 // to get an extended Kalman filter functionality.

video.h

@@ -15,6 +15,7 @@ typedef cv::Ptr<cv::BackgroundSubtractorKNN>* BackgroundSubtractorKNN;
 typedef cv::Ptr<cv::Tracker>* Tracker;
 typedef cv::Ptr<cv::TrackerMIL>* TrackerMIL;
 typedef cv::Ptr<cv::TrackerGOTURN>* TrackerGOTURN;
+typedef cv::Ptr<cv::TrackerVit>* TrackerVit;
 typedef cv::KalmanFilter* KalmanFilter;
 #else
 typedef void* BackgroundSubtractorMOG2;
@@ -22,6 +23,7 @@ typedef void* BackgroundSubtractorKNN;
 typedef void* Tracker;
 typedef void* TrackerMIL;
 typedef void* TrackerGOTURN;
+typedef void* TrackerVit;
 typedef void* KalmanFilter;
 #endif
 
@@ -54,6 +56,10 @@ TrackerGOTURN TrackerGOTURN_Create(void);
 TrackerGOTURN TrackerGOTURN_CreateWithParams(const char* modelBin, const char* modelTxt);
 void TrackerGOTURN_Close(TrackerGOTURN tr);
 
+TrackerVit TrackerVit_Create();
+TrackerVit TrackerVit_CreateWithParams(const char* model);
+void TrackerVit_Close(TrackerVit self);
+
 KalmanFilter KalmanFilter_New(int dynamParams, int measureParams);
 KalmanFilter KalmanFilter_NewWithParams(int dynamParams, int measureParams, int controlParams, int type);
 void KalmanFilter_Close(KalmanFilter kf);

video_test.go

@@ -334,12 +334,18 @@ func TestSingleTrackers(t *testing.T) {
 		goturnPath = "./testdata"
 	}
 
+	vitPath := os.Getenv("GOCV_ONNX_TEST_FILES")
+	if vitPath == "" {
+		vitPath = "./testdata"
+	}
+
 	tab := []struct {
 		name    string
 		tracker Tracker
 	}{
 		{"MIL", NewTrackerMIL()},
 		{"GOTURN", NewTrackerGOTURNWithParams(goturnPath+"/goturn.caffemodel", goturnPath+"/goturn.prototxt")},
+		{"Vit", NewTrackerVitWithParams(vitPath + "/object_tracking_vittrack_2023sep.onnx")},
 	}
 
 	for _, test := range tab {