Support Python binding for CUDA functionalities

modules/python/test/test_cuda.py

@@ -0,0 +1,45 @@
+#!/usr/bin/env python
+
+'''
+CUDA-accelerated Computer Vision functions
+'''
+
+# Python 2/3 compatibility
+from __future__ import print_function
+
+import numpy as np
+import cv2 as cv
+
+from tests_common import NewOpenCVTests
+
+class cuda_test(NewOpenCVTests):
+    def setUp(self):
+        if not cv.cuda.getCudaEnabledDeviceCount():
+            self.skipTest("No CUDA-capable device is detected")
+
+    def test_cuda_upload_download(self):
+        npMat = (np.random.random((200, 200, 3)) * 255).astype(np.uint8)
+        gpuMat = cv.cuda_GpuMat()
+        gpuMat.upload(npMat)
+
+        self.assertTrue(np.allclose(gpuMat.download(), npMat))
+
+    def test_cuda_imgproc_cvtColor(self):
+        npMat = (np.random.random((200, 200, 3)) * 255).astype(np.uint8)
+        gpuMat = cv.cuda_GpuMat()
+        gpuMat.upload(npMat)
+        gpuMat2 = cv.cuda.cvtColor(gpuMat, cv.COLOR_BGR2HSV)
+
+        self.assertTrue(np.allclose(gpuMat2.download(), cv.cvtColor(npMat, cv.COLOR_BGR2HSV)))
+
+    def test_cuda_filter_laplacian(self):
+        npMat = (np.random.random((200, 200)) * 255).astype(np.uint16)
+        gpuMat = cv.cuda_GpuMat()
+        gpuMat.upload(npMat)
+        gpuMat = cv.cuda.createLaplacianFilter(cv.CV_16UC1, -1, ksize=3).apply(gpuMat)
+
+        self.assertTrue(np.allclose(gpuMat.download(), cv.Laplacian(npMat, cv.CV_16UC1, ksize=3)))
+
+
+if __name__ == '__main__':
+    NewOpenCVTests.bootstrap()