Merge pull request #26875 from asmorkalov:as/in_memory_models

Added trackers factory with pre-loaded dnn models #26875

Replaces https://github.com/opencv/opencv/pull/26295

Allows to substitute custom models or initialize tracker from in-memory model.

### Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

- [x] I agree to contribute to the project under Apache 2 License.
- [x] To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
- [ ] The PR is proposed to the proper branch
- [ ] There is a reference to the original bug report and related work
- [ ] There is accuracy test, performance test and test data in opencv_extra repository, if applicable
      Patch to opencv_extra has the same branch name.
- [ ] The feature is well documented and sample code can be built with the project CMake

modules/java/generator/gen_java.py

@@ -425,6 +425,8 @@ class FuncInfo(GeneralInfo):
             arg_fix_map = func_fix_map.get(arg[1], {})
             arg[0] = arg_fix_map.get('ctype',  arg[0]) #fixing arg type
             arg[3] = arg_fix_map.get('attrib', arg[3]) #fixing arg attrib
+            if arg[0] == 'dnn_Net':
+                arg[0] = 'Net'
             self.args.append(ArgInfo(arg))
 
     def fullClassJAVA(self):
@@ -474,7 +476,7 @@ class JavaWrapperGenerator(object):
             jni_name = "(*("+classinfo.fullNameCPP()+"*)%(n)s_nativeObj)"
         type_dict.setdefault(name, {}).update(
             { "j_type" : classinfo.jname,
-              "jn_type" : "long", "jn_args" : (("__int64", ".nativeObj"),),
+              "jn_type" : "long", "jn_args" : (("__int64", ".getNativeObjAddr()"),),
               "jni_name" : jni_name,
               "jni_type" : "jlong",
               "suffix" : "J",
@@ -483,7 +485,7 @@ class JavaWrapperGenerator(object):
         )
         type_dict.setdefault(name+'*', {}).update(
             { "j_type" : classinfo.jname,
-              "jn_type" : "long", "jn_args" : (("__int64", ".nativeObj"),),
+              "jn_type" : "long", "jn_args" : (("__int64", ".getNativeObjAddr()"),),
               "jni_name" : "&("+jni_name+")",
               "jni_type" : "jlong",
               "suffix" : "J",

modules/python/test/tests_common.py

@@ -24,11 +24,12 @@ class NewOpenCVTests(unittest.TestCase):
     # path to local repository folder containing 'samples' folder
     repoPath = None
     extraTestDataPath = None
+    extraDnnTestDataPath = None
     # github repository url
     repoUrl = 'https://raw.github.com/opencv/opencv/4.x'
 
     def find_file(self, filename, searchPaths=[], required=True):
-        searchPaths = searchPaths if searchPaths else [self.repoPath, self.extraTestDataPath]
+        searchPaths = searchPaths if searchPaths else [self.repoPath, self.extraTestDataPath, self.extraDnnTestDataPath]
         for path in searchPaths:
             if path is not None:
                 candidate = path + '/' + filename
@@ -83,10 +84,17 @@ class NewOpenCVTests(unittest.TestCase):
         print("Testing OpenCV", cv.__version__)
         print("Local repo path:", args.repo)
         NewOpenCVTests.repoPath = args.repo
+
         try:
             NewOpenCVTests.extraTestDataPath = os.environ['OPENCV_TEST_DATA_PATH']
         except KeyError:
             print('Missing opencv extra repository. Some of tests may fail.')
+
+        try:
+            NewOpenCVTests.extraDnnTestDataPath = os.environ['OPENCV_DNN_TEST_DATA_PATH']
+        except KeyError:
+            pass
+
         random.seed(0)
         unit_argv = [sys.argv[0]] + other
         unittest.main(argv=unit_argv)

modules/video/include/opencv2/video/tracking.hpp

@@ -46,6 +46,9 @@
 
 #include "opencv2/core.hpp"
 #include "opencv2/imgproc.hpp"
+#ifdef HAVE_OPENCV_DNN
+# include "opencv2/dnn.hpp"
+#endif
 
 namespace cv
 {
@@ -826,6 +829,13 @@ public:
     static CV_WRAP
     Ptr<TrackerGOTURN> create(const TrackerGOTURN::Params& parameters = TrackerGOTURN::Params());
 
+#ifdef HAVE_OPENCV_DNN
+    /** @brief Constructor
+    @param model pre-loaded GOTURN model
+    */
+    static CV_WRAP Ptr<TrackerGOTURN> create(const dnn::Net& model);
+#endif
+
     //void init(InputArray image, const Rect& boundingBox) CV_OVERRIDE;
     //bool update(InputArray image, CV_OUT Rect& boundingBox) CV_OVERRIDE;
 };
@@ -853,6 +863,16 @@ public:
     static CV_WRAP
     Ptr<TrackerDaSiamRPN> create(const TrackerDaSiamRPN::Params& parameters = TrackerDaSiamRPN::Params());
 
+#ifdef HAVE_OPENCV_DNN
+    /** @brief Constructor
+     *  @param siam_rpn pre-loaded SiamRPN model
+     *  @param kernel_cls1 pre-loaded CLS model
+     *  @param kernel_r1 pre-loaded R1 model
+     */
+    static CV_WRAP
+    Ptr<TrackerDaSiamRPN> create(const dnn::Net& siam_rpn, const dnn::Net& kernel_cls1, const dnn::Net& kernel_r1);
+#endif
+
     /** @brief Return tracking score
     */
     CV_WRAP virtual float getTrackingScore() = 0;
@@ -891,6 +911,15 @@ public:
     static CV_WRAP
     Ptr<TrackerNano> create(const TrackerNano::Params& parameters = TrackerNano::Params());
 
+#ifdef HAVE_OPENCV_DNN
+    /** @brief Constructor
+     *  @param backbone pre-loaded backbone model
+     *  @param neckhead pre-loaded neckhead model
+     */
+    static CV_WRAP
+    Ptr<TrackerNano> create(const dnn::Net& backbone, const dnn::Net& neckhead);
+#endif
+
     /** @brief Return tracking score
     */
     CV_WRAP virtual float getTrackingScore() = 0;
@@ -929,6 +958,18 @@ public:
     static CV_WRAP
     Ptr<TrackerVit> create(const TrackerVit::Params& parameters = TrackerVit::Params());
 
+#ifdef HAVE_OPENCV_DNN
+    /** @brief Constructor
+     *  @param model pre-loaded DNN model
+     *  @param meanvalue mean value for image preprocessing
+     *  @param stdvalue std value for image preprocessing
+     *  @param tracking_score_threshold threshold for tracking score
+     */
+    static CV_WRAP
+    Ptr<TrackerVit> create(const dnn::Net& model, Scalar meanvalue = Scalar(0.485, 0.456, 0.406),
+                           Scalar stdvalue = Scalar(0.229, 0.224, 0.225), float tracking_score_threshold = 0.20f);
+#endif
+
     /** @brief Return tracking score
     */
     CV_WRAP virtual float getTrackingScore() = 0;

modules/video/misc/java/test/TrackerCreateTest.java

@@ -1,31 +1,98 @@
 package org.opencv.test.video;
 
+import java.io.File;
 import org.opencv.core.Core;
 import org.opencv.core.CvType;
 import org.opencv.core.CvException;
 import org.opencv.core.Mat;
 import org.opencv.core.Rect;
+import org.opencv.dnn.Dnn;
+import org.opencv.dnn.Net;
 import org.opencv.test.OpenCVTestCase;
 
 import org.opencv.video.Tracker;
 import org.opencv.video.TrackerGOTURN;
+import org.opencv.video.TrackerGOTURN_Params;
+import org.opencv.video.TrackerNano;
+import org.opencv.video.TrackerNano_Params;
+import org.opencv.video.TrackerVit;
+import org.opencv.video.TrackerVit_Params;
 import org.opencv.video.TrackerMIL;
 
 public class TrackerCreateTest extends OpenCVTestCase {
 
+    private final static String ENV_OPENCV_DNN_TEST_DATA_PATH = "OPENCV_DNN_TEST_DATA_PATH";
+    private final static String ENV_OPENCV_TEST_DATA_PATH = "OPENCV_TEST_DATA_PATH";
+    private String testDataPath;
+    private String modelsDataPath;
+
     @Override
     protected void setUp() throws Exception {
         super.setUp();
+
+        // relys on https://developer.android.com/reference/java/lang/System
+        isTestCaseEnabled = System.getProperties().getProperty("java.vm.name") != "Dalvik";
+        if (!isTestCaseEnabled) {
+            return;
+        }
+
+        testDataPath = System.getenv(ENV_OPENCV_TEST_DATA_PATH);
+        if (testDataPath == null) {
+            throw new Exception(ENV_OPENCV_TEST_DATA_PATH + " has to be defined!");
+        }
+
+        modelsDataPath = System.getenv(ENV_OPENCV_DNN_TEST_DATA_PATH);
+        if (modelsDataPath == null) {
+            modelsDataPath = testDataPath;
+        }
+
+        if (isTestCaseEnabled) {
+            testDataPath = System.getenv(ENV_OPENCV_DNN_TEST_DATA_PATH);
+            if (testDataPath == null)
+                testDataPath = System.getenv(ENV_OPENCV_TEST_DATA_PATH);
+            if (testDataPath == null)
+                throw new Exception(ENV_OPENCV_TEST_DATA_PATH + " has to be defined!");
+        }
     }
 
-
     public void testCreateTrackerGOTURN() {
+        Net net;
         try {
-            Tracker tracker = TrackerGOTURN.create();
-            assert(tracker != null);
+            String protoFile = new File(testDataPath, "dnn/gsoc2016-goturn/goturn.prototxt").toString();
+            String weightsFile = new File(modelsDataPath, "dnn/gsoc2016-goturn/goturn.caffemodel").toString();
+            net = Dnn.readNetFromCaffe(protoFile, weightsFile);
         } catch (CvException e) {
-            // expected, model files may be missing
+            return;
         }
+        Tracker tracker = TrackerGOTURN.create(net);
+        assert(tracker != null);
+    }
+
+    public void testCreateTrackerNano() {
+        Net backbone;
+        Net neckhead;
+        try {
+            String backboneFile = new File(modelsDataPath, "dnn/onnx/models/nanotrack_backbone_sim_v2.onnx").toString();
+            String neckheadFile = new File(modelsDataPath, "dnn/onnx/models/nanotrack_head_sim_v2.onnx").toString();
+            backbone = Dnn.readNet(backboneFile);
+            neckhead = Dnn.readNet(neckheadFile);
+        } catch (CvException e) {
+            return;
+        }
+        Tracker tracker = TrackerNano.create(backbone, neckhead);
+        assert(tracker != null);
+    }
+
+    public void testCreateTrackerVit() {
+        Net net;
+        try {
+            String backboneFile = new File(modelsDataPath, "dnn/onnx/models/vitTracker.onnx").toString();
+            net = Dnn.readNet(backboneFile);
+        } catch (CvException e) {
+            return;
+        }
+        Tracker tracker = TrackerVit.create(net);
+        assert(tracker != null);
     }
 
     public void testCreateTrackerMIL() {
@@ -35,5 +102,4 @@ public class TrackerCreateTest extends OpenCVTestCase {
         Rect rect = new Rect(10, 10, 30, 30);
         tracker.init(mat, rect);  // should not crash (https://github.com/opencv/opencv/issues/19915)
     }
-
 }

modules/video/misc/python/test/test_tracking.py

@@ -7,12 +7,30 @@ from tests_common import NewOpenCVTests, unittest
 
 class tracking_test(NewOpenCVTests):
 
-    def test_createTracker(self):
-        t = cv.TrackerMIL_create()
-        try:
-            t = cv.TrackerGOTURN_create()
-        except cv.error as e:
-            pass  # may fail due to missing DL model files
+    def test_createMILTracker(self):
+        t = cv.TrackerMIL.create()
+        self.assertTrue(t is not None)
+
+    def test_createGoturnTracker(self):
+        proto = self.find_file("dnn/gsoc2016-goturn/goturn.prototxt", required=False);
+        weights = self.find_file("dnn/gsoc2016-goturn/goturn.caffemodel", required=False);
+        net = cv.dnn.readNet(proto, weights)
+        t = cv.TrackerGOTURN.create(net)
+        self.assertTrue(t is not None)
+
+    def test_createNanoTracker(self):
+        backbone_path = self.find_file("dnn/onnx/models/nanotrack_backbone_sim_v2.onnx", required=False);
+        neckhead_path = self.find_file("dnn/onnx/models/nanotrack_head_sim_v2.onnx", required=False);
+        backbone = cv.dnn.readNet(backbone_path)
+        neckhead = cv.dnn.readNet(neckhead_path)
+        t = cv.TrackerNano.create(backbone, neckhead)
+        self.assertTrue(t is not None)
+
+    def test_createVitTracker(self):
+        model_path = self.find_file("dnn/onnx/models/vitTracker.onnx", required=False);
+        model = cv.dnn.readNet(model_path)
+        t = cv.TrackerVit.create(model)
+        self.assertTrue(t is not None)
 
 
 if __name__ == '__main__':

modules/video/src/tracking/tracker_dasiamrpn.cpp

@@ -57,10 +57,8 @@ Mat sizeCal(const Mat& w, const Mat& h)
 class TrackerDaSiamRPNImpl : public TrackerDaSiamRPN
 {
 public:
-    TrackerDaSiamRPNImpl(const TrackerDaSiamRPN::Params& parameters)
-        : params(parameters)
+    TrackerDaSiamRPNImpl(const TrackerDaSiamRPN::Params& params)
     {
-
         siamRPN = dnn::readNet(params.model);
         siamKernelCL1 = dnn::readNet(params.kernel_cls1);
         siamKernelR1 = dnn::readNet(params.kernel_r1);
@@ -77,12 +75,21 @@ public:
         siamKernelCL1.setPreferableTarget(params.target);
     }
 
+    TrackerDaSiamRPNImpl(const dnn::Net& siam_rpn, const dnn::Net& kernel_cls1, const dnn::Net& kernel_r1)
+    {
+        CV_Assert(!siam_rpn.empty());
+        CV_Assert(!kernel_cls1.empty());
+        CV_Assert(!kernel_r1.empty());
+
+        siamRPN = siam_rpn;
+        siamKernelCL1 = kernel_cls1;
+        siamKernelR1 = kernel_r1;
+    }
+
     void init(InputArray image, const Rect& boundingBox) CV_OVERRIDE;
     bool update(InputArray image, Rect& boundingBox) CV_OVERRIDE;
     float getTrackingScore() CV_OVERRIDE;
 
-    TrackerDaSiamRPN::Params params;
-
 protected:
     dnn::Net siamRPN, siamKernelR1, siamKernelCL1;
     Rect boundingBox_;
@@ -425,16 +432,22 @@ Mat TrackerDaSiamRPNImpl::getSubwindow(Mat& img, const Rect2f& targetBox, float
 
     return zCrop;
 }
+
 Ptr<TrackerDaSiamRPN> TrackerDaSiamRPN::create(const TrackerDaSiamRPN::Params& parameters)
 {
     return makePtr<TrackerDaSiamRPNImpl>(parameters);
 }
 
+Ptr<TrackerDaSiamRPN> TrackerDaSiamRPN::create(const dnn::Net& siam_rpn, const dnn::Net& kernel_cls1, const dnn::Net& kernel_r1)
+{
+    return makePtr<TrackerDaSiamRPNImpl>(siam_rpn, kernel_cls1, kernel_r1);
+}
+
 #else  // OPENCV_HAVE_DNN
 Ptr<TrackerDaSiamRPN> TrackerDaSiamRPN::create(const TrackerDaSiamRPN::Params& parameters)
 {
     (void)(parameters);
-    CV_Error(cv::Error::StsNotImplemented, "to use GOTURN, the tracking module needs to be built with opencv_dnn !");
+    CV_Error(cv::Error::StsNotImplemented, "to use DaSiamRPN, the tracking module needs to be built with opencv_dnn !");
 }
 #endif  // OPENCV_HAVE_DNN
 }

modules/video/src/tracking/tracker_goturn.cpp

@@ -31,11 +31,16 @@ TrackerGOTURN::Params::Params()
 class TrackerGOTURNImpl : public TrackerGOTURN
 {
 public:
+    TrackerGOTURNImpl(const dnn::Net& model)
+    {
+        CV_Assert(!model.empty());
+        net = model;
+    }
+
     TrackerGOTURNImpl(const TrackerGOTURN::Params& parameters)
-        : params(parameters)
     {
         // Load GOTURN architecture from *.prototxt and pretrained weights from *.caffemodel
-        net = dnn::readNetFromCaffe(params.modelTxt, params.modelBin);
+        net = dnn::readNetFromCaffe(parameters.modelTxt, parameters.modelBin);
         CV_Assert(!net.empty());
     }
 
@@ -49,8 +54,6 @@ public:
         boundingBox_ = boundingBox & Rect(Point(0, 0), image_.size());
     }
 
-    TrackerGOTURN::Params params;
-
     dnn::Net net;
     Rect boundingBox_;
     Mat image_;
@@ -129,6 +132,11 @@ Ptr<TrackerGOTURN> TrackerGOTURN::create(const TrackerGOTURN::Params& parameters
     return makePtr<TrackerGOTURNImpl>(parameters);
 }
 
+Ptr<TrackerGOTURN> TrackerGOTURN::create(const dnn::Net& model)
+{
+    return makePtr<TrackerGOTURNImpl>(model);
+}
+
 #else  // OPENCV_HAVE_DNN
 Ptr<TrackerGOTURN> TrackerGOTURN::create(const TrackerGOTURN::Params& parameters)
 {

modules/video/src/tracking/tracker_nano.cpp

@@ -87,18 +87,26 @@ class TrackerNanoImpl : public TrackerNano
 {
 public:
     TrackerNanoImpl(const TrackerNano::Params& parameters)
-        : params(parameters)
     {
-        backbone = dnn::readNet(params.backbone);
-        neckhead = dnn::readNet(params.neckhead);
+        backbone = dnn::readNet(parameters.backbone);
+        neckhead = dnn::readNet(parameters.neckhead);
 
         CV_Assert(!backbone.empty());
         CV_Assert(!neckhead.empty());
 
-        backbone.setPreferableBackend(params.backend);
-        backbone.setPreferableTarget(params.target);
-        neckhead.setPreferableBackend(params.backend);
-        neckhead.setPreferableTarget(params.target);
+        backbone.setPreferableBackend(parameters.backend);
+        backbone.setPreferableTarget(parameters.target);
+        neckhead.setPreferableBackend(parameters.backend);
+        neckhead.setPreferableTarget(parameters.target);
+    }
+
+    TrackerNanoImpl(const dnn::Net& _backbone, const dnn::Net& _neckhead)
+    {
+        CV_Assert(!_backbone.empty());
+        CV_Assert(!_neckhead.empty());
+
+        backbone = _backbone;
+        neckhead = _neckhead;
     }
 
     void init(InputArray image, const Rect& boundingBox) CV_OVERRIDE;
@@ -110,8 +118,6 @@ public:
     std::vector<float> targetPos = {0, 0}; // center point of bounding box (x, y)
     float tracking_score;
 
-    TrackerNano::Params params;
-
     struct trackerConfig
     {
         float windowInfluence = 0.455f;
@@ -349,6 +355,11 @@ Ptr<TrackerNano> TrackerNano::create(const TrackerNano::Params& parameters)
     return makePtr<TrackerNanoImpl>(parameters);
 }
 
+Ptr<TrackerNano> TrackerNano::create(const dnn::Net& backbone, const dnn::Net& neckhead)
+{
+    return makePtr<TrackerNanoImpl>(backbone, neckhead);
+}
+
 #else  // OPENCV_HAVE_DNN
 Ptr<TrackerNano> TrackerNano::create(const TrackerNano::Params& parameters)
 {

modules/video/src/tracking/tracker_vit.cpp

@@ -42,16 +42,26 @@ class TrackerVitImpl : public TrackerVit
 {
 public:
     TrackerVitImpl(const TrackerVit::Params& parameters)
-        : params(parameters)
     {
-        net = dnn::readNet(params.net);
+        net = dnn::readNet(parameters.net);
         CV_Assert(!net.empty());
 
-        net.setPreferableBackend(params.backend);
-        net.setPreferableTarget(params.target);
+        net.setPreferableBackend(parameters.backend);
+        net.setPreferableTarget(parameters.target);
 
-        i2bp.mean = params.meanvalue * 255.0;
-        i2bp.scalefactor = (1.0 / params.stdvalue) * (1 / 255.0);
+        i2bp.mean = parameters.meanvalue * 255.0;
+        i2bp.scalefactor = (1.0 / parameters.stdvalue) * (1 / 255.0);
+        tracking_score_threshold = parameters.tracking_score_threshold;
+    }
+
+    TrackerVitImpl(const dnn::Net& model, Scalar meanvalue, Scalar stdvalue, float _tracking_score_threshold)
+    {
+        CV_Assert(!model.empty());
+
+        net = model;
+        i2bp.mean = meanvalue * 255.0;
+        i2bp.scalefactor = (1.0 / stdvalue) * (1 / 255.0);
+        tracking_score_threshold = _tracking_score_threshold;
     }
 
     void init(InputArray image, const Rect& boundingBox) CV_OVERRIDE;
@@ -61,7 +71,7 @@ public:
     Rect rect_last;
     float tracking_score;
 
-    TrackerVit::Params params;
+    float tracking_score_threshold;
     dnn::Image2BlobParams i2bp;
 
 
@@ -189,7 +199,7 @@ bool TrackerVitImpl::update(InputArray image_, Rect &boundingBoxRes)
     minMaxLoc(conf_map, nullptr, &maxVal, nullptr, &maxLoc);
     tracking_score = static_cast<float>(maxVal);
 
-    if (tracking_score >= params.tracking_score_threshold) {
+    if (tracking_score >= tracking_score_threshold) {
         float cx = (maxLoc.x + offset_map.at<float>(0, maxLoc.y, maxLoc.x)) / 16;
         float cy = (maxLoc.y + offset_map.at<float>(1, maxLoc.y, maxLoc.x)) / 16;
         float w = size_map.at<float>(0, maxLoc.y, maxLoc.x);
@@ -213,6 +223,11 @@ Ptr<TrackerVit> TrackerVit::create(const TrackerVit::Params& parameters)
     return makePtr<TrackerVitImpl>(parameters);
 }
 
+Ptr<TrackerVit> TrackerVit::create(const dnn::Net& model, Scalar meanvalue, Scalar stdvalue, float tracking_score_threshold)
+{
+    return makePtr<TrackerVitImpl>(model, meanvalue, stdvalue, tracking_score_threshold);
+}
+
 #else  // OPENCV_HAVE_DNN
 Ptr<TrackerVit> TrackerVit::create(const TrackerVit::Params& parameters)
 {