Merge pull request #2396 from cudawarped:fix_python_cudawarping_cudaarithm

Add python bindings to cudaobjdetect, cudawarping and cudaarithm

* Overload cudawarping functions to generate correct python bindings.
Add python wrapper to convolution funciton.

* Added shift and hog.

* Moved cuda python tests to this repo and added python bindings to SURF.

* Fix SURF documentation and allow meanshiftsegmention to create GpuMat internaly if not passed for python bindings consistency.

* Add correct cuda SURF test case.

* Fix python mog and mog2 python bindings, add tests and  correct cudawarping documentation.

* Updated KeyPoints in cuda::ORB::Convert python wrapper to be an output argument.

* Add changes suggested by alalek

* Added changes suggested by asmorkalov

modules/cudaarithm/include/opencv2/cudaarithm.hpp

@@ -274,6 +274,10 @@ CV_EXPORTS_W void bitwise_xor(InputArray src1, InputArray src2, OutputArray dst,
  */
 CV_EXPORTS void rshift(InputArray src, Scalar_<int> val, OutputArray dst, Stream& stream = Stream::Null());
 
+CV_WRAP inline void rshift(InputArray src, Scalar val, OutputArray dst, Stream& stream = Stream::Null()) {
+    rshift(src, Scalar_<int>(val), dst, stream);
+}
+
 /** @brief Performs pixel by pixel right left of an image by a constant value.
 
 @param src Source matrix. Supports 1, 3 and 4 channels images with CV_8U , CV_16U or CV_32S
@@ -284,6 +288,10 @@ depth.
  */
 CV_EXPORTS void lshift(InputArray src, Scalar_<int> val, OutputArray dst, Stream& stream = Stream::Null());
 
+CV_WRAP inline void lshift(InputArray src, Scalar val, OutputArray dst, Stream& stream = Stream::Null()) {
+    lshift(src, Scalar_<int>(val), dst, stream);
+}
+
 /** @brief Computes the per-element minimum of two matrices (or a matrix and a scalar).
 
 @param src1 First source matrix or scalar.
@@ -858,7 +866,7 @@ public:
     @param ccorr Flags to evaluate cross-correlation instead of convolution.
     @param stream Stream for the asynchronous version.
      */
-    virtual void convolve(InputArray image, InputArray templ, OutputArray result, bool ccorr = false, Stream& stream = Stream::Null()) = 0;
+    CV_WRAP virtual void convolve(InputArray image, InputArray templ, OutputArray result, bool ccorr = false, Stream& stream = Stream::Null()) = 0;
 };
 
 /** @brief Creates implementation for cuda::Convolution .

modules/cudaarithm/misc/python/test/test_cudaarithm.py

@@ -0,0 +1,178 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudaarithm_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudaarithm_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_cudaarithm(self):
+        npMat = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
+
+        cuMat = cv.cuda_GpuMat(npMat)
+        cuMatDst = cv.cuda_GpuMat(cuMat.size(),cuMat.type())
+        cuMatB = cv.cuda_GpuMat(cuMat.size(),cv.CV_8UC1)
+        cuMatG = cv.cuda_GpuMat(cuMat.size(),cv.CV_8UC1)
+        cuMatR = cv.cuda_GpuMat(cuMat.size(),cv.CV_8UC1)
+
+        self.assertTrue(np.allclose(cv.cuda.merge(cv.cuda.split(cuMat)),npMat))
+
+        cv.cuda.split(cuMat,[cuMatB,cuMatG,cuMatR])
+        cv.cuda.merge([cuMatB,cuMatG,cuMatR],cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),npMat))
+
+        shift = (np.random.random((cuMat.channels(),)) * 8).astype(np.uint8).tolist()
+        self.assertTrue(np.allclose(cv.cuda.rshift(cuMat,shift).download(),npMat  >> shift))
+        cv.cuda.rshift(cuMat,shift,cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),npMat >> shift))
+
+        self.assertTrue(np.allclose(cv.cuda.lshift(cuMat,shift).download(),(npMat << shift).astype('uint8')))
+        cv.cuda.lshift(cuMat,shift,cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),(npMat << shift).astype('uint8')))
+
+    def test_arithmetic(self):
+        npMat1 = np.random.random((128, 128, 3)) - 0.5
+        npMat2 = np.random.random((128, 128, 3)) - 0.5
+
+        cuMat1 = cv.cuda_GpuMat()
+        cuMat2 = cv.cuda_GpuMat()
+        cuMat1.upload(npMat1)
+        cuMat2.upload(npMat2)
+        cuMatDst = cv.cuda_GpuMat(cuMat1.size(),cuMat1.type())
+
+        self.assertTrue(np.allclose(cv.cuda.add(cuMat1, cuMat2).download(),
+                                         cv.add(npMat1, npMat2)))
+
+        cv.cuda.add(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.add(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.subtract(cuMat1, cuMat2).download(),
+                                         cv.subtract(npMat1, npMat2)))
+
+        cv.cuda.subtract(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.subtract(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.multiply(cuMat1, cuMat2).download(),
+                                         cv.multiply(npMat1, npMat2)))
+
+        cv.cuda.multiply(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.multiply(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.divide(cuMat1, cuMat2).download(),
+                                         cv.divide(npMat1, npMat2)))
+
+        cv.cuda.divide(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.divide(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.absdiff(cuMat1, cuMat2).download(),
+                                         cv.absdiff(npMat1, npMat2)))
+
+        cv.cuda.absdiff(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.absdiff(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.compare(cuMat1, cuMat2, cv.CMP_GE).download(),
+                                         cv.compare(npMat1, npMat2, cv.CMP_GE)))
+
+        cuMatDst1 = cv.cuda_GpuMat(cuMat1.size(),cv.CV_8UC3)
+        cv.cuda.compare(cuMat1, cuMat2, cv.CMP_GE, cuMatDst1)
+        self.assertTrue(np.allclose(cuMatDst1.download(),cv.compare(npMat1, npMat2, cv.CMP_GE)))
+
+        self.assertTrue(np.allclose(cv.cuda.abs(cuMat1).download(),
+                                         np.abs(npMat1)))
+
+        cv.cuda.abs(cuMat1, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),np.abs(npMat1)))
+
+        self.assertTrue(np.allclose(cv.cuda.sqrt(cv.cuda.sqr(cuMat1)).download(),
+                                    cv.cuda.abs(cuMat1).download()))
+
+        cv.cuda.sqr(cuMat1, cuMatDst)
+        cv.cuda.sqrt(cuMatDst, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.cuda.abs(cuMat1).download()))
+
+        self.assertTrue(np.allclose(cv.cuda.log(cv.cuda.exp(cuMat1)).download(),
+                                                            npMat1))
+
+        cv.cuda.exp(cuMat1, cuMatDst)
+        cv.cuda.log(cuMatDst, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),npMat1))
+
+        self.assertTrue(np.allclose(cv.cuda.pow(cuMat1, 2).download(),
+                                         cv.pow(npMat1, 2)))
+
+        cv.cuda.pow(cuMat1, 2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.pow(npMat1, 2)))
+
+    def test_logical(self):
+        npMat1 = (np.random.random((128, 128)) * 255).astype(np.uint8)
+        npMat2 = (np.random.random((128, 128)) * 255).astype(np.uint8)
+
+        cuMat1 = cv.cuda_GpuMat()
+        cuMat2 = cv.cuda_GpuMat()
+        cuMat1.upload(npMat1)
+        cuMat2.upload(npMat2)
+        cuMatDst = cv.cuda_GpuMat(cuMat1.size(),cuMat1.type())
+
+        self.assertTrue(np.allclose(cv.cuda.bitwise_or(cuMat1, cuMat2).download(),
+                                         cv.bitwise_or(npMat1, npMat2)))
+
+        cv.cuda.bitwise_or(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.bitwise_or(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.bitwise_and(cuMat1, cuMat2).download(),
+                                         cv.bitwise_and(npMat1, npMat2)))
+
+        cv.cuda.bitwise_and(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.bitwise_and(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.bitwise_xor(cuMat1, cuMat2).download(),
+                                         cv.bitwise_xor(npMat1, npMat2)))
+
+        cv.cuda.bitwise_xor(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.bitwise_xor(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.bitwise_not(cuMat1).download(),
+                                         cv.bitwise_not(npMat1)))
+
+        cv.cuda.bitwise_not(cuMat1, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.bitwise_not(npMat1)))
+
+        self.assertTrue(np.allclose(cv.cuda.min(cuMat1, cuMat2).download(),
+                                         cv.min(npMat1, npMat2)))
+
+        cv.cuda.min(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.min(npMat1, npMat2)))
+
+        self.assertTrue(np.allclose(cv.cuda.max(cuMat1, cuMat2).download(),
+                                         cv.max(npMat1, npMat2)))
+
+        cv.cuda.max(cuMat1, cuMat2, cuMatDst)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.max(npMat1, npMat2)))
+
+    def test_convolution(self):
+        npMat = (np.random.random((128, 128)) * 255).astype(np.float32)
+        npDims = np.array(npMat.shape)
+        kernel = (np.random.random((3, 3)) * 1).astype(np.float32)
+        kernelDims = np.array(kernel.shape)
+        iS = (kernelDims/2).astype(int)
+        iE = npDims - kernelDims + iS
+
+        cuMat = cv.cuda_GpuMat(npMat)
+        cuKernel= cv.cuda_GpuMat(kernel)
+        cuMatDst = cv.cuda_GpuMat(tuple(npDims - kernelDims + 1), cuMat.type())
+        conv = cv.cuda.createConvolution()
+
+        self.assertTrue(np.allclose(conv.convolve(cuMat,cuKernel,ccorr=True).download(),
+                    cv.filter2D(npMat,-1,kernel,anchor=(-1,-1))[iS[0]:iE[0]+1,iS[1]:iE[1]+1]))
+
+        conv.convolve(cuMat,cuKernel,cuMatDst,True)
+        self.assertTrue(np.allclose(cuMatDst.download(),
+                    cv.filter2D(npMat,-1,kernel,anchor=(-1,-1))[iS[0]:iE[0]+1,iS[1]:iE[1]+1]))
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudabgsegm/include/opencv2/cudabgsegm.hpp

@@ -85,7 +85,11 @@ public:
     CV_WRAP virtual void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream) = 0;
 
     using cv::BackgroundSubtractor::getBackgroundImage;
-    CV_WRAP virtual void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const = 0;
+    virtual void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const = 0;
+
+    CV_WRAP inline void getBackgroundImage(CV_OUT GpuMat& backgroundImage, Stream& stream) {
+        getBackgroundImage(OutputArray(backgroundImage), stream);
+    }
 
     CV_WRAP virtual int getHistory() const = 0;
     CV_WRAP virtual void setHistory(int nframes) = 0;
@@ -131,7 +135,11 @@ public:
 
     CV_WRAP virtual void apply(InputArray image, OutputArray fgmask, double learningRate, Stream& stream) = 0;
 
-    CV_WRAP virtual void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const = 0;
+    virtual void getBackgroundImage(OutputArray backgroundImage, Stream& stream) const = 0;
+
+    CV_WRAP inline void getBackgroundImage(CV_OUT GpuMat &backgroundImage, Stream& stream) {
+        getBackgroundImage(OutputArray(backgroundImage), stream);
+    }
 };
 
 /** @brief Creates MOG2 Background Subtractor

modules/cudabgsegm/misc/python/test/test_cudabgsegm.py

@@ -0,0 +1,43 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudabgsegm_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudabgsegm_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_cudabgsegm(self):
+        lr = 0.05
+        sz = (128,128,1)
+        npMat = (np.random.random(sz) * 255).astype(np.uint8)
+        cuMat = cv.cuda_GpuMat(npMat)
+        cuMatBg = cv.cuda_GpuMat(cuMat.size(),cuMat.type())
+        cuMatFg = cv.cuda_GpuMat(cuMat.size(),cuMat.type())
+
+        mog = cv.cuda.createBackgroundSubtractorMOG()
+        mog.apply(cuMat, lr, cv.cuda.Stream_Null(), cuMatFg)
+        mog.getBackgroundImage(cv.cuda.Stream_Null(),cuMatBg)
+        self.assertTrue(sz[:2] == cuMatFg.size() == cuMatBg.size())
+        self.assertTrue(sz[2] == cuMatFg.channels() == cuMatBg.channels())
+        self.assertTrue(cv.CV_8UC1 == cuMatFg.type() == cuMatBg.type())
+        mog = cv.cuda.createBackgroundSubtractorMOG()
+        self.assertTrue(np.allclose(cuMatFg.download(),mog.apply(cuMat, lr, cv.cuda.Stream_Null()).download()))
+        self.assertTrue(np.allclose(cuMatBg.download(),mog.getBackgroundImage(cv.cuda.Stream_Null()).download()))
+
+        mog2 = cv.cuda.createBackgroundSubtractorMOG2()
+        mog2.apply(cuMat, lr, cv.cuda.Stream_Null(), cuMatFg)
+        mog2.getBackgroundImage(cv.cuda.Stream_Null(),cuMatBg)
+        self.assertTrue(sz[:2] == cuMatFg.size() == cuMatBg.size())
+        self.assertTrue(sz[2] == cuMatFg.channels() == cuMatBg.channels())
+        self.assertTrue(cv.CV_8UC1 == cuMatFg.type() == cuMatBg.type())
+        mog2 = cv.cuda.createBackgroundSubtractorMOG2()
+        self.assertTrue(np.allclose(cuMatFg.download(),mog2.apply(cuMat, lr, cv.cuda.Stream_Null()).download()))
+        self.assertTrue(np.allclose(cuMatBg.download(),mog2.getBackgroundImage(cv.cuda.Stream_Null()).download()))
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudacodec/misc/python/test/test_cudacodec.py

@@ -0,0 +1,55 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudacodec_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudacodec_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    @unittest.skipIf('OPENCV_TEST_DATA_PATH' not in os.environ,
+                     "OPENCV_TEST_DATA_PATH is not defined")
+    def test_reader(self):
+        #Test the functionality but not the results of the video reader
+
+        vid_path = os.environ['OPENCV_TEST_DATA_PATH'] + '/cv/video/1920x1080.avi'
+        try:
+            reader = cv.cudacodec.createVideoReader(vid_path)
+            ret, gpu_mat = reader.nextFrame()
+            self.assertTrue(ret)
+            self.assertTrue('GpuMat' in str(type(gpu_mat)), msg=type(gpu_mat))
+            #TODO: print(cv.utils.dumpInputArray(gpu_mat)) # - no support for GpuMat
+
+            # not checking output, therefore sepearate tests for different signatures is unecessary
+            ret, _gpu_mat2 = reader.nextFrame(gpu_mat)
+            #TODO: self.assertTrue(gpu_mat == gpu_mat2)
+            self.assertTrue(ret)
+        except cv.error as e:
+            notSupported = (e.code == cv.Error.StsNotImplemented or e.code == cv.Error.StsUnsupportedFormat or e.code == cv.Error.GPU_API_CALL_ERROR)
+            self.assertTrue(notSupported)
+            if e.code == cv.Error.StsNotImplemented:
+                self.skipTest("NVCUVID is not installed")
+            elif e.code == cv.Error.StsUnsupportedFormat:
+                self.skipTest("GPU hardware video decoder missing or video format not supported")
+            elif e.code == cv.Error.GPU_API_CALL_ERRROR:
+                self.skipTest("GPU hardware video decoder is missing")
+            else:
+                self.skipTest(e.err)
+
+    def test_writer_existence(self):
+        #Test at least the existence of wrapped functions for now
+
+        try:
+            _writer = cv.cudacodec.createVideoWriter("tmp", (128, 128), 30)
+        except cv.error as e:
+            self.assertEqual(e.code, cv.Error.StsNotImplemented)
+            self.skipTest("NVCUVENC is not installed")
+
+        self.assertTrue(True) #It is sufficient that no exceptions have been there
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudafeatures2d/include/opencv2/cudafeatures2d.hpp

@@ -414,7 +414,7 @@ public:
 
     /** Converts keypoints array from internal representation to standard vector. */
     CV_WRAP virtual void convert(InputArray gpu_keypoints,
-                         std::vector<KeyPoint>& keypoints) = 0;
+                         CV_OUT std::vector<KeyPoint>& keypoints) = 0;
 };
 
 //

modules/cudafeatures2d/misc/python/test/test_cudafeatures2d.py

@@ -0,0 +1,47 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudafeatures2d_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudafeatures2d_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_cudafeatures2d(self):
+        npMat1 = self.get_sample("samples/data/right01.jpg")
+        npMat2 = self.get_sample("samples/data/right02.jpg")
+
+        cuMat1 = cv.cuda_GpuMat()
+        cuMat2 = cv.cuda_GpuMat()
+        cuMat1.upload(npMat1)
+        cuMat2.upload(npMat2)
+
+        cuMat1 = cv.cuda.cvtColor(cuMat1, cv.COLOR_RGB2GRAY)
+        cuMat2 = cv.cuda.cvtColor(cuMat2, cv.COLOR_RGB2GRAY)
+
+        fast = cv.cuda_FastFeatureDetector.create()
+        _kps = fast.detectAsync(cuMat1)
+
+        orb = cv.cuda_ORB.create()
+        _kps1, descs1 = orb.detectAndComputeAsync(cuMat1, None)
+        _kps2, descs2 = orb.detectAndComputeAsync(cuMat2, None)
+
+        self.assertTrue(len(orb.convert(_kps1)) == _kps1.size()[0])
+        self.assertTrue(len(orb.convert(_kps2)) == _kps2.size()[0])
+
+        bf = cv.cuda_DescriptorMatcher.createBFMatcher(cv.NORM_HAMMING)
+        matches = bf.match(descs1, descs2)
+        self.assertGreater(len(matches), 0)
+        matches = bf.knnMatch(descs1, descs2, 2)
+        self.assertGreater(len(matches), 0)
+        matches = bf.radiusMatch(descs1, descs2, 0.1)
+        self.assertGreater(len(matches), 0)
+
+        self.assertTrue(True) #It is sufficient that no exceptions have been there
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudafilters/misc/python/test/test_cudafilters.py

@@ -0,0 +1,43 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudafilters_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudafilters_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_existence(self):
+        #Test at least the existence of wrapped functions for now
+
+        _filter = cv.cuda.createBoxFilter(cv.CV_8UC1, -1, (3, 3))
+        _filter = cv.cuda.createLinearFilter(cv.CV_8UC4, -1, np.eye(3))
+        _filter = cv.cuda.createLaplacianFilter(cv.CV_16UC1, -1, ksize=3)
+        _filter = cv.cuda.createSeparableLinearFilter(cv.CV_8UC1, -1, np.eye(3), np.eye(3))
+        _filter = cv.cuda.createDerivFilter(cv.CV_8UC1, -1, 1, 1, 3)
+        _filter = cv.cuda.createSobelFilter(cv.CV_8UC1, -1, 1, 1)
+        _filter = cv.cuda.createScharrFilter(cv.CV_8UC1, -1, 1, 0)
+        _filter = cv.cuda.createGaussianFilter(cv.CV_8UC1, -1, (3, 3), 16)
+        _filter = cv.cuda.createMorphologyFilter(cv.MORPH_DILATE, cv.CV_32FC1, np.eye(3))
+        _filter = cv.cuda.createBoxMaxFilter(cv.CV_8UC1, (3, 3))
+        _filter = cv.cuda.createBoxMinFilter(cv.CV_8UC1, (3, 3))
+        _filter = cv.cuda.createRowSumFilter(cv.CV_8UC1, cv.CV_32FC1, 3)
+        _filter = cv.cuda.createColumnSumFilter(cv.CV_8UC1, cv.CV_32FC1, 3)
+        _filter = cv.cuda.createMedianFilter(cv.CV_8UC1, 3)
+
+        self.assertTrue(True) #It is sufficient that no exceptions have been there
+
+    def test_laplacian(self):
+        npMat = (np.random.random((128, 128)) * 255).astype(np.uint16)
+        cuMat = cv.cuda_GpuMat()
+        cuMat.upload(npMat)
+
+        self.assertTrue(np.allclose(cv.cuda.createLaplacianFilter(cv.CV_16UC1, -1, ksize=3).apply(cuMat).download(),
+                                         cv.Laplacian(npMat, cv.CV_16UC1, ksize=3)))
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudaimgproc/misc/python/test/test_cudaimgproc.py

@@ -0,0 +1,93 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudaimgproc_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudaimgproc_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_cudaimgproc(self):
+        npC1 = (np.random.random((128, 128)) * 255).astype(np.uint8)
+        npC3 = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
+        npC4 = (np.random.random((128, 128, 4)) * 255).astype(np.uint8)
+        cuC1 = cv.cuda_GpuMat()
+        cuC3 = cv.cuda_GpuMat()
+        cuC4 = cv.cuda_GpuMat()
+        cuC1.upload(npC1)
+        cuC3.upload(npC3)
+        cuC4.upload(npC4)
+
+        cv.cuda.cvtColor(cuC3, cv.COLOR_RGB2HSV)
+        cv.cuda.demosaicing(cuC1, cv.cuda.COLOR_BayerGR2BGR_MHT)
+        cv.cuda.gammaCorrection(cuC3)
+        cv.cuda.alphaComp(cuC4, cuC4, cv.cuda.ALPHA_XOR)
+        cv.cuda.calcHist(cuC1)
+        cv.cuda.equalizeHist(cuC1)
+        cv.cuda.evenLevels(3, 0, 255)
+        cv.cuda.meanShiftFiltering(cuC4, 10, 5)
+        cv.cuda.meanShiftProc(cuC4, 10, 5)
+        cv.cuda.bilateralFilter(cuC3, 3, 16, 3)
+        cv.cuda.blendLinear
+
+        cuRes = cv.cuda.meanShiftSegmentation(cuC4, 10, 5, 5)
+        cuDst = cv.cuda_GpuMat(cuC4.size(),cuC4.type())
+        cv.cuda.meanShiftSegmentation(cuC4, 10, 5, 5, cuDst)
+        self.assertTrue(np.allclose(cuRes.download(),cuDst.download()))
+
+        clahe = cv.cuda.createCLAHE()
+        clahe.apply(cuC1, cv.cuda_Stream.Null())
+
+        histLevels = cv.cuda.histEven(cuC3, 20, 0, 255)
+        cv.cuda.histRange(cuC1, histLevels)
+
+        detector = cv.cuda.createCannyEdgeDetector(0, 100)
+        detector.detect(cuC1)
+
+        detector = cv.cuda.createHoughLinesDetector(3, np.pi / 180, 20)
+        detector.detect(cuC1)
+
+        detector = cv.cuda.createHoughSegmentDetector(3, np.pi / 180, 20, 5)
+        detector.detect(cuC1)
+
+        detector = cv.cuda.createHoughCirclesDetector(3, 20, 10, 10, 20, 100)
+        detector.detect(cuC1)
+
+        detector = cv.cuda.createGeneralizedHoughBallard()
+        #BUG: detect accept only Mat!
+        #Even if generate_gpumat_decls is set to True, it only wraps overload CUDA functions.
+        #The problem is that Mat and GpuMat are not fully compatible to enable system-wide overloading
+        #detector.detect(cuC1, cuC1, cuC1)
+
+        detector = cv.cuda.createGeneralizedHoughGuil()
+        #BUG: same as above..
+        #detector.detect(cuC1, cuC1, cuC1)
+
+        detector = cv.cuda.createHarrisCorner(cv.CV_8UC1, 15, 5, 1)
+        detector.compute(cuC1)
+
+        detector = cv.cuda.createMinEigenValCorner(cv.CV_8UC1, 15, 5, 1)
+        detector.compute(cuC1)
+
+        detector = cv.cuda.createGoodFeaturesToTrackDetector(cv.CV_8UC1)
+        detector.detect(cuC1)
+
+        matcher = cv.cuda.createTemplateMatching(cv.CV_8UC1, cv.TM_CCOEFF_NORMED)
+        matcher.match(cuC3, cuC3)
+
+        self.assertTrue(True) #It is sufficient that no exceptions have been there
+
+    def test_cvtColor(self):
+        npMat = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
+        cuMat = cv.cuda_GpuMat()
+        cuMat.upload(npMat)
+
+        self.assertTrue(np.allclose(cv.cuda.cvtColor(cuMat, cv.COLOR_BGR2HSV).download(),
+                                         cv.cvtColor(npMat, cv.COLOR_BGR2HSV)))
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudaimgproc/src/mssegmentation.cpp

@@ -388,7 +388,15 @@ void cv::cuda::meanShiftSegmentation(InputArray _src, OutputArray _dst, int sp,
             dstcol[3] = 255;
         }
     }
+
+    if (_dst.kind() == _InputArray::CUDA_GPU_MAT)
+    {
+        GpuMat dstGpuMat = getOutputMat(_dst, src.size(), src.type(), stream);
+        dstGpuMat.upload(dst, stream);
+    }
+    else {
         dst.copyTo(_dst);
     }
+}
 
 #endif // #if !defined (HAVE_CUDA) || defined (CUDA_DISABLER)

modules/cudaobjdetect/include/opencv2/cudaobjdetect.hpp

@@ -162,6 +162,22 @@ public:
                         std::vector<Point>& found_locations,
                         std::vector<double>* confidences = NULL) = 0;
 
+    CV_WRAP inline void detect(InputArray img,
+        CV_OUT std::vector<Point>& found_locations,
+        CV_OUT std::vector<double>& confidences) {
+        detect(img, found_locations, &confidences);
+    }
+
+    /** @brief Performs object detection without a multi-scale window.
+
+    @param img Source image. CV_8UC1 and CV_8UC4 types are supported for now.
+    @param found_locations Left-top corner points of detected objects boundaries.
+     */
+    CV_WRAP inline void detectWithoutConf(InputArray img,
+        CV_OUT std::vector<Point>& found_locations) {
+        detect(img, found_locations, NULL);
+    }
+
     /** @brief Performs object detection with a multi-scale window.
 
     @param img Source image. See cuda::HOGDescriptor::detect for type limitations.
@@ -172,6 +188,22 @@ public:
                                   std::vector<Rect>& found_locations,
                                   std::vector<double>* confidences = NULL) = 0;
 
+    CV_WRAP inline void detectMultiScale(InputArray img,
+        CV_OUT std::vector<Rect>& found_locations,
+        CV_OUT std::vector<double>& confidences) {
+        detectMultiScale(img, found_locations, &confidences);
+    }
+
+    /** @brief Performs object detection with a multi-scale window.
+
+    @param img Source image. See cuda::HOGDescriptor::detect for type limitations.
+    @param found_locations Detected objects boundaries.
+     */
+    CV_WRAP inline void detectMultiScaleWithoutConf(InputArray img,
+        CV_OUT std::vector<Rect>& found_locations) {
+        detectMultiScale(img, found_locations, NULL);
+    }
+
     /** @brief Returns block descriptors computed for the whole image.
 
     @param img Source image. See cuda::HOGDescriptor::detect for type limitations.

modules/cudaobjdetect/misc/python/test/test_cudaobjdetect.py

@@ -0,0 +1,38 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class cudaobjdetect_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudaobjdetect_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    @unittest.skipIf('OPENCV_TEST_DATA_PATH' not in os.environ,
+                        "OPENCV_TEST_DATA_PATH is not defined")
+    def test_hog(self):
+        img_path = os.environ['OPENCV_TEST_DATA_PATH'] + '/gpu/caltech/image_00000009_0.png'
+        npMat = cv.cvtColor(cv.imread(img_path),cv.COLOR_BGR2BGRA)
+
+        cuMat = cv.cuda_GpuMat(npMat)
+        cuHog = cv.cuda.HOG_create()
+        cuHog.setSVMDetector(cuHog.getDefaultPeopleDetector())
+
+        loc, conf = cuHog.detect(cuMat)
+        self.assertTrue(len(loc) == len(conf) and len(loc) > 0 and len(loc[0]) == 2)
+
+        loc = cuHog.detectWithoutConf(cuMat)
+        self.assertTrue(len(loc) > 0 and len(loc[0]) == 2)
+
+        loc = cuHog.detectMultiScaleWithoutConf(cuMat)
+        self.assertTrue(len(loc) > 0 and len(loc[0]) == 4)
+
+        cuHog.setGroupThreshold(0)
+        loc, conf = cuHog.detectMultiScale(cuMat)
+        self.assertTrue(len(loc) == len(conf) and len(loc) > 0 and len(loc[0]) == 4)
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/cudawarping/include/opencv2/cudawarping.hpp

@@ -112,7 +112,7 @@ CV_EXPORTS_W void resize(InputArray src, OutputArray dst, Size dsize, double fx=
 @param src Source image. CV_8U , CV_16U , CV_32S , or CV_32F depth and 1, 3, or 4 channels are
 supported.
 @param dst Destination image with the same type as src . The size is dsize .
-@param M *2x3* transformation matrix.
+@param M *2x3* Mat or UMat transformation matrix.
 @param dsize Size of the destination image.
 @param flags Combination of interpolation methods (see resize) and the optional flag
 WARP_INVERSE_MAP specifying that M is an inverse transformation ( dst=\>src ). Only
@@ -123,12 +123,22 @@ INTER_NEAREST , INTER_LINEAR , and INTER_CUBIC interpolation methods are support
 
 @sa warpAffine
  */
-CV_EXPORTS_W void warpAffine(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
+CV_EXPORTS void warpAffine(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
     int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
 
+CV_WRAP inline void warpAffine(InputArray src, OutputArray dst, UMat M, Size dsize, int flags = INTER_LINEAR,
+    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null()) {
+    warpAffine(src, dst, InputArray(M), dsize, flags, borderMode, borderValue, stream);
+}
+
+CV_WRAP inline void warpAffine(InputArray src, OutputArray dst, Mat M, Size dsize, int flags = INTER_LINEAR,
+    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null()) {
+    warpAffine(src, dst, InputArray(M), dsize, flags, borderMode, borderValue, stream);
+}
+
 /** @brief Builds transformation maps for affine transformation.
 
-@param M *2x3* transformation matrix.
+@param M *2x3* Mat or UMat transformation matrix.
 @param inverse Flag specifying that M is an inverse transformation ( dst=\>src ).
 @param dsize Size of the destination image.
 @param xmap X values with CV_32FC1 type.
@@ -137,14 +147,22 @@ CV_EXPORTS_W void warpAffine(InputArray src, OutputArray dst, InputArray M, Size
 
 @sa cuda::warpAffine , cuda::remap
  */
-CV_EXPORTS_W void buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
+CV_EXPORTS void buildWarpAffineMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
+
+CV_WRAP inline void buildWarpAffineMaps(UMat M, bool inverse, Size dsize, CV_OUT GpuMat& xmap, CV_OUT GpuMat& ymap, Stream& stream = Stream::Null()) {
+    buildWarpAffineMaps(InputArray(M), inverse, dsize, OutputArray(xmap), OutputArray(ymap), stream);
+}
+
+CV_WRAP inline void buildWarpAffineMaps(Mat M, bool inverse, Size dsize, CV_OUT GpuMat& xmap, CV_OUT GpuMat& ymap, Stream& stream = Stream::Null()) {
+    buildWarpAffineMaps(InputArray(M), inverse, dsize, OutputArray(xmap), OutputArray(ymap), stream);
+}
 
 /** @brief Applies a perspective transformation to an image.
 
 @param src Source image. CV_8U , CV_16U , CV_32S , or CV_32F depth and 1, 3, or 4 channels are
 supported.
 @param dst Destination image with the same type as src . The size is dsize .
-@param M *3x3* transformation matrix.
+@param M *3x3* Mat or UMat transformation matrix.
 @param dsize Size of the destination image.
 @param flags Combination of interpolation methods (see resize ) and the optional flag
 WARP_INVERSE_MAP specifying that M is the inverse transformation ( dst =\> src ). Only
@@ -155,12 +173,22 @@ INTER_NEAREST , INTER_LINEAR , and INTER_CUBIC interpolation methods are support
 
 @sa warpPerspective
  */
-CV_EXPORTS_W void warpPerspective(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
+CV_EXPORTS void warpPerspective(InputArray src, OutputArray dst, InputArray M, Size dsize, int flags = INTER_LINEAR,
     int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null());
 
+CV_WRAP inline void warpPerspective(InputArray src, OutputArray dst, UMat M, Size dsize, int flags = INTER_LINEAR,
+    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null()) {
+    warpPerspective(src, dst, InputArray(M), dsize, flags, borderMode, borderValue, stream);
+}
+
+CV_WRAP inline void warpPerspective(InputArray src, OutputArray dst, Mat M, Size dsize, int flags = INTER_LINEAR,
+    int borderMode = BORDER_CONSTANT, Scalar borderValue = Scalar(), Stream& stream = Stream::Null()) {
+    warpPerspective(src, dst, InputArray(M), dsize, flags, borderMode, borderValue, stream);
+}
+
 /** @brief Builds transformation maps for perspective transformation.
 
-@param M *3x3* transformation matrix.
+@param M *3x3* Mat or UMat transformation matrix.
 @param inverse Flag specifying that M is an inverse transformation ( dst=\>src ).
 @param dsize Size of the destination image.
 @param xmap X values with CV_32FC1 type.
@@ -169,7 +197,15 @@ CV_EXPORTS_W void warpPerspective(InputArray src, OutputArray dst, InputArray M,
 
 @sa cuda::warpPerspective , cuda::remap
  */
-CV_EXPORTS_W void buildWarpPerspectiveMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
+CV_EXPORTS void buildWarpPerspectiveMaps(InputArray M, bool inverse, Size dsize, OutputArray xmap, OutputArray ymap, Stream& stream = Stream::Null());
+
+CV_WRAP inline void buildWarpPerspectiveMaps(UMat M, bool inverse, Size dsize, CV_OUT GpuMat& xmap, CV_OUT GpuMat& ymap, Stream& stream = Stream::Null()) {
+    buildWarpPerspectiveMaps(InputArray(M), inverse, dsize, OutputArray(xmap), OutputArray(ymap), stream);
+}
+
+CV_WRAP inline void buildWarpPerspectiveMaps(Mat M, bool inverse, Size dsize, CV_OUT GpuMat& xmap, CV_OUT GpuMat& ymap, Stream& stream = Stream::Null()) {
+    buildWarpPerspectiveMaps(InputArray(M), inverse, dsize, OutputArray(xmap), OutputArray(ymap), stream);
+}
 
 /** @brief Rotates an image around the origin (0,0) and then shifts it.
 

modules/cudawarping/misc/python/test/test_cudawarping.py

@@ -0,0 +1,76 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+def create_affine_transform_matrix(size,angle):
+    return np.array([[np.cos(angle), -np.sin(angle), size[1]/2], [np.sin(angle), np.cos(angle), 0]])
+def create_perspective_transform_matrix(size,angle):
+    return np.vstack([create_affine_transform_matrix(size,angle),[0, 0, 1]])
+
+class cudawarping_test(NewOpenCVTests):
+    def setUp(self):
+        super(cudawarping_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_resize(self):
+        dstSz = (256,256)
+        interp = cv.INTER_NEAREST
+        npMat = (np.random.random((128,128,3))*255).astype(np.uint8)
+
+        cuMat = cv.cuda_GpuMat(npMat)
+        cuMatDst = cv.cuda_GpuMat(dstSz,cuMat.type())
+
+        self.assertTrue(np.allclose(cv.cuda.resize(cuMat,dstSz,interpolation=interp).download(),
+            cv.resize(npMat,dstSz,interpolation=interp)))
+
+        cv.cuda.resize(cuMat,dstSz,cuMatDst,interpolation=interp)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.resize(npMat,dstSz,interpolation=interp)))
+
+    def test_warp(self):
+        npMat = (np.random.random((128,128,3))*255).astype(np.uint8)
+        size = npMat.shape[:2]
+        M1 = create_affine_transform_matrix(size,np.pi/2)
+
+        cuMat = cv.cuda_GpuMat(npMat)
+        cuMatDst = cv.cuda_GpuMat(size,cuMat.type())
+
+        borderType = cv.BORDER_REFLECT101
+        self.assertTrue(np.allclose(cv.cuda.warpAffine(cuMat,M1,size,borderMode=borderType).download(),
+            cv.warpAffine(npMat,M1,size, borderMode=borderType)))
+        cv.cuda.warpAffine(cuMat,M1,size,cuMatDst,borderMode=borderType)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.warpAffine(npMat,M1,size,borderMode=borderType)))
+
+        interpolation = cv.INTER_NEAREST
+        flags = interpolation | cv.WARP_INVERSE_MAP
+        dst_gold = cv.warpAffine(npMat, M1, size, flags = flags)
+        cuMaps = cv.cuda.buildWarpAffineMaps(M1,True,size)
+        dst = cv.remap(npMat, cuMaps[0].download(), cuMaps[1].download(),interpolation)
+        self.assertTrue(np.allclose(dst,dst_gold))
+
+        xmap = cv.cuda_GpuMat(size,cv.CV_32FC1)
+        ymap = cv.cuda_GpuMat(size,cv.CV_32FC1)
+        cv.cuda.buildWarpAffineMaps(M1,True,size,xmap,ymap)
+        dst = cv.remap(npMat, xmap.download(), ymap.download(),interpolation)
+        self.assertTrue(np.allclose(dst,dst_gold))
+
+        M2 = create_perspective_transform_matrix(size,np.pi/2)
+        np.allclose(cv.cuda.warpPerspective(cuMat,M2,size,borderMode=borderType).download(),
+                    cv.warpPerspective(npMat,M2,size,borderMode=borderType))
+        cv.cuda.warpPerspective(cuMat,M2,size,cuMatDst,borderMode=borderType)
+        self.assertTrue(np.allclose(cuMatDst.download(),cv.warpPerspective(npMat,M2,size,borderMode=borderType)))
+
+        dst_gold = cv.warpPerspective(npMat, M2, size, flags = flags)
+        cuMaps = cv.cuda.buildWarpPerspectiveMaps(M2,True,size)
+        dst = cv.remap(npMat, cuMaps[0].download(), cuMaps[1].download(),interpolation)
+        self.assertTrue(np.allclose(dst,dst_gold))
+
+        cv.cuda.buildWarpPerspectiveMaps(M2,True,size,xmap,ymap)
+        dst = cv.remap(npMat, xmap.download(), ymap.download(),interpolation)
+        self.assertTrue(np.allclose(dst,dst_gold))
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/xfeatures2d/include/opencv2/xfeatures2d/cuda.hpp

@@ -83,7 +83,7 @@ between function calls.
         opencv_source_code/samples/gpu/surf_keypoint_matcher.cpp
 
  */
-class CV_EXPORTS SURF_CUDA
+class CV_EXPORTS_W SURF_CUDA
 {
 public:
     enum KeypointLayout
@@ -104,15 +104,28 @@ public:
     explicit SURF_CUDA(double _hessianThreshold, int _nOctaves=4,
          int _nOctaveLayers=2, bool _extended=false, float _keypointsRatio=0.01f, bool _upright = false);
 
+    /**
+    @param _hessianThreshold Threshold for hessian keypoint detector used in SURF.
+    @param _nOctaves Number of pyramid octaves the keypoint detector will use.
+    @param _nOctaveLayers Number of octave layers within each octave.
+    @param _extended Extended descriptor flag (true - use extended 128-element descriptors; false - use
+    64-element descriptors).
+    @param _keypointsRatio
+    @param _upright Up-right or rotated features flag (true - do not compute orientation of features;
+    false - compute orientation).
+    */
+    CV_WRAP static Ptr<SURF_CUDA> create(double _hessianThreshold, int _nOctaves = 4,
+        int _nOctaveLayers = 2, bool _extended = false, float _keypointsRatio = 0.01f, bool _upright = false);
+
     //! returns the descriptor size in float's (64 or 128)
-    int descriptorSize() const;
+    CV_WRAP int descriptorSize() const;
     //! returns the default norm type
-    int defaultNorm() const;
+    CV_WRAP int defaultNorm() const;
 
     //! upload host keypoints to device memory
     void uploadKeypoints(const std::vector<KeyPoint>& keypoints, GpuMat& keypointsGPU);
     //! download keypoints from device to host memory
-    void downloadKeypoints(const GpuMat& keypointsGPU, std::vector<KeyPoint>& keypoints);
+    CV_WRAP void downloadKeypoints(const GpuMat& keypointsGPU, CV_OUT std::vector<KeyPoint>& keypoints);
 
     //! download descriptors from device to host memory
     void downloadDescriptors(const GpuMat& descriptorsGPU, std::vector<float>& descriptors);
@@ -133,24 +146,47 @@ public:
     void operator()(const GpuMat& img, const GpuMat& mask, GpuMat& keypoints, GpuMat& descriptors,
         bool useProvidedKeypoints = false);
 
+    /** @brief Finds the keypoints using fast hessian detector used in SURF
+
+    @param img Source image, currently supports only CV_8UC1 images.
+    @param mask A mask image same size as src and of type CV_8UC1.
+    @param keypoints Detected keypoints.
+     */
+    CV_WRAP inline void detect(const GpuMat& img, const GpuMat& mask, CV_OUT GpuMat& keypoints) {
+        (*this)(img, mask, keypoints);
+    }
+
     void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints);
     void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, GpuMat& descriptors,
         bool useProvidedKeypoints = false);
 
+    /** @brief Finds the keypoints and computes their descriptors using fast hessian detector used in SURF
+
+    @param img Source image, currently supports only CV_8UC1 images.
+    @param mask A mask image same size as src and of type CV_8UC1.
+    @param keypoints Detected keypoints.
+    @param descriptors Keypoint descriptors.
+    @param useProvidedKeypoints Compute descriptors for the user-provided keypoints and recompute keypoints direction.
+     */
+    CV_WRAP inline void detectWithDescriptors(const GpuMat& img, const GpuMat& mask, CV_OUT GpuMat& keypoints, CV_OUT GpuMat& descriptors,
+        bool useProvidedKeypoints = false) {
+        (*this)(img, mask, keypoints, descriptors, useProvidedKeypoints);
+    }
+
     void operator()(const GpuMat& img, const GpuMat& mask, std::vector<KeyPoint>& keypoints, std::vector<float>& descriptors,
         bool useProvidedKeypoints = false);
 
     void releaseMemory();
 
     // SURF parameters
-    double hessianThreshold;
-    int nOctaves;
-    int nOctaveLayers;
-    bool extended;
-    bool upright;
+    CV_PROP double hessianThreshold;
+    CV_PROP int nOctaves;
+    CV_PROP int nOctaveLayers;
+    CV_PROP bool extended;
+    CV_PROP bool upright;
 
     //! max keypoints = min(keypointsRatio * img.size().area(), 65535)
-    float keypointsRatio;
+    CV_PROP float keypointsRatio;
 
     GpuMat sum, mask1, maskSum;
 

modules/xfeatures2d/misc/python/test/test_cuda_xfeatures2d.py

@@ -0,0 +1,48 @@
+#!/usr/bin/env python
+import os
+import cv2 as cv
+import numpy as np
+
+from tests_common import NewOpenCVTests, unittest
+
+class xfeatures2d_test(NewOpenCVTests):
+    def setUp(self):
+        super(xfeatures2d_test, self).setUp()
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    @unittest.skipIf('OPENCV_TEST_DATA_PATH' not in os.environ,
+                     "OPENCV_TEST_DATA_PATH is not defined")
+    def test_surf(self):
+        img_path = os.environ['OPENCV_TEST_DATA_PATH'] + "/gpu/features2d/aloe.png"
+        hessianThreshold = 100
+        nOctaves = 3
+        nOctaveLayers = 2
+        extended = False
+        keypointsRatio = 0.05
+        upright = False
+
+        npMat = cv.cvtColor(cv.imread(img_path),cv.COLOR_BGR2GRAY)
+        cuMat = cv.cuda_GpuMat(npMat)
+
+        try:
+            cuSurf = cv.cuda_SURF_CUDA.create(hessianThreshold,nOctaves,nOctaveLayers,extended,keypointsRatio,upright)
+            surf = cv.xfeatures2d_SURF.create(hessianThreshold,nOctaves,nOctaveLayers,extended,upright)
+        except cv.error as e:
+            self.assertEqual(e.code, cv.Error.StsNotImplemented)
+            self.skipTest("OPENCV_ENABLE_NONFREE is not enabled in this build.")
+
+        cuKeypoints = cuSurf.detect(cuMat,cv.cuda_GpuMat())
+        keypointsHost = cuSurf.downloadKeypoints(cuKeypoints)
+        keypoints = surf.detect(npMat)
+        self.assertTrue(len(keypointsHost) == len(keypoints))
+
+        cuKeypoints, cuDescriptors = cuSurf.detectWithDescriptors(cuMat,cv.cuda_GpuMat(),cuKeypoints,useProvidedKeypoints=True)
+        keypointsHost = cuSurf.downloadKeypoints(cuKeypoints)
+        descriptorsHost = cuDescriptors.download()
+        keypoints, descriptors = surf.compute(npMat,keypoints)
+
+        self.assertTrue(len(keypointsHost) == len(keypoints) and descriptorsHost.shape == descriptors.shape)
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()

modules/xfeatures2d/src/surf.cuda.cpp

@@ -54,6 +54,7 @@ using namespace cv::cuda;
 
 cv::cuda::SURF_CUDA::SURF_CUDA() { throw_no_nonfree }
 cv::cuda::SURF_CUDA::SURF_CUDA(double, int, int, bool, float, bool) { throw_no_nonfree }
+Ptr<SURF_CUDA> cv::cuda::SURF_CUDA::create(double, int, int, bool, float, bool) { throw_no_nonfree }
 int cv::cuda::SURF_CUDA::descriptorSize() const { throw_no_nonfree }
 int cv::cuda::SURF_CUDA::defaultNorm() const { throw_no_nonfree }
 void cv::cuda::SURF_CUDA::uploadKeypoints(const std::vector<KeyPoint>&, GpuMat&) { throw_no_nonfree }
@@ -71,6 +72,7 @@ void cv::cuda::SURF_CUDA::releaseMemory() { throw_no_nonfree }
 
 cv::cuda::SURF_CUDA::SURF_CUDA() { throw_no_cuda(); }
 cv::cuda::SURF_CUDA::SURF_CUDA(double, int, int, bool, float, bool) { throw_no_cuda(); }
+Ptr<SURF_CUDA> cv::cuda::SURF_CUDA::create(double, int, int, bool, float, bool) { throw_no_cuda(); }
 int cv::cuda::SURF_CUDA::descriptorSize() const { throw_no_cuda(); }
 int cv::cuda::SURF_CUDA::defaultNorm() const { throw_no_cuda(); }
 void cv::cuda::SURF_CUDA::uploadKeypoints(const std::vector<KeyPoint>&, GpuMat&) { throw_no_cuda(); }
@@ -451,6 +453,12 @@ void cv::cuda::SURF_CUDA::releaseMemory()
     maxPosBuffer.release();
 }
 
+Ptr<SURF_CUDA> cv::cuda::SURF_CUDA::create(double _hessianThreshold, int _nOctaves,
+    int _nOctaveLayers, bool _extended, float _keypointsRatio, bool _upright)
+{
+    return makePtr<SURF_CUDA>(_hessianThreshold, _nOctaves, _nOctaveLayers, _extended, _keypointsRatio, _upright);
+}
+
 #endif // !defined (OPENCV_ENABLE_NONFREE)
 #endif