1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
|
#!/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
import os
from tests_common import NewOpenCVTests, unittest
class cuda_test(NewOpenCVTests):
def setUp(self):
super(cuda_test, self).setUp()
if not cv.cuda.getCudaEnabledDeviceCount():
self.skipTest("No CUDA-capable device is detected")
def test_cuda_upload_download(self):
npMat = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
cuMat = cv.cuda_GpuMat()
cuMat.upload(npMat)
self.assertTrue(np.allclose(cuMat.download(), npMat))
def test_cuda_upload_download_stream(self):
stream = cv.cuda_Stream()
npMat = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
cuMat = cv.cuda_GpuMat(128,128, cv.CV_8UC3)
cuMat.upload(npMat, stream)
npMat2 = cuMat.download(stream=stream)
stream.waitForCompletion()
self.assertTrue(np.allclose(npMat2, npMat))
def test_cuda_interop(self):
npMat = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
cuMat = cv.cuda_GpuMat()
cuMat.upload(npMat)
self.assertTrue(cuMat.cudaPtr() != 0)
cuMatFromPtrSz = cv.cuda.createGpuMatFromCudaMemory(cuMat.size(),cuMat.type(),cuMat.cudaPtr(), cuMat.step)
self.assertTrue(cuMat.cudaPtr() == cuMatFromPtrSz.cudaPtr())
cuMatFromPtrRc = cv.cuda.createGpuMatFromCudaMemory(cuMat.size()[1],cuMat.size()[0],cuMat.type(),cuMat.cudaPtr(), cuMat.step)
self.assertTrue(cuMat.cudaPtr() == cuMatFromPtrRc.cudaPtr())
stream = cv.cuda_Stream()
self.assertTrue(stream.cudaPtr() != 0)
streamFromPtr = cv.cuda.wrapStream(stream.cudaPtr())
self.assertTrue(stream.cudaPtr() == streamFromPtr.cudaPtr())
asyncstream = cv.cuda_Stream(1) # cudaStreamNonBlocking
self.assertTrue(asyncstream.cudaPtr() != 0)
def test_cuda_buffer_pool(self):
cv.cuda.setBufferPoolUsage(True)
cv.cuda.setBufferPoolConfig(cv.cuda.getDevice(), 1024 * 1024 * 64, 2)
stream_a = cv.cuda.Stream()
pool_a = cv.cuda.BufferPool(stream_a)
cuMat = pool_a.getBuffer(1024, 1024, cv.CV_8UC3)
cv.cuda.setBufferPoolUsage(False)
self.assertEqual(cuMat.size(), (1024, 1024))
self.assertEqual(cuMat.type(), cv.CV_8UC3)
def test_cuda_release(self):
npMat = (np.random.random((128, 128, 3)) * 255).astype(np.uint8)
cuMat = cv.cuda_GpuMat()
cuMat.upload(npMat)
cuMat.release()
self.assertTrue(cuMat.cudaPtr() == 0)
self.assertTrue(cuMat.step == 0)
self.assertTrue(cuMat.size() == (0, 0))
def test_cuda_convertTo(self):
# setup
npMat_8UC4 = (np.random.random((128, 128, 4)) * 255).astype(np.uint8)
npMat_32FC4 = npMat_8UC4.astype(np.single)
new_type = cv.CV_32FC4
# sync
# in/out
cuMat_8UC4 = cv.cuda_GpuMat(npMat_8UC4)
cuMat_32FC4 = cv.cuda_GpuMat(cuMat_8UC4.size(), new_type)
cuMat_32FC4_out = cuMat_8UC4.convertTo(new_type, cuMat_32FC4)
self.assertTrue(cuMat_32FC4.cudaPtr() == cuMat_32FC4_out.cudaPtr())
npMat_32FC4_out = cuMat_32FC4.download()
self.assertTrue(np.array_equal(npMat_32FC4, npMat_32FC4_out))
# out
cuMat_32FC4_out = cuMat_8UC4.convertTo(new_type)
npMat_32FC4_out = cuMat_32FC4.download()
self.assertTrue(np.array_equal(npMat_32FC4, npMat_32FC4_out))
# async
stream = cv.cuda.Stream()
cuMat_32FC4 = cv.cuda_GpuMat(cuMat_8UC4.size(), new_type)
cuMat_32FC4_out = cuMat_8UC4.convertTo(new_type, cuMat_32FC4)
# in/out
cuMat_32FC4_out = cuMat_8UC4.convertTo(new_type, 1, 0, stream, cuMat_32FC4)
self.assertTrue(cuMat_32FC4.cudaPtr() == cuMat_32FC4_out.cudaPtr())
npMat_32FC4_out = cuMat_32FC4.download(stream)
stream.waitForCompletion()
self.assertTrue(np.array_equal(npMat_32FC4, npMat_32FC4_out))
# out
cuMat_32FC4_out = cuMat_8UC4.convertTo(new_type, 1, 0, stream)
npMat_32FC4_out = cuMat_32FC4.download(stream)
stream.waitForCompletion()
self.assertTrue(np.array_equal(npMat_32FC4, npMat_32FC4_out))
def test_cuda_copyTo(self):
# setup
npMat_8UC4 = (np.random.random((128, 128, 4)) * 255).astype(np.uint8)
# sync
# in/out
cuMat_8UC4 = cv.cuda_GpuMat(npMat_8UC4)
cuMat_8UC4_dst = cv.cuda_GpuMat(cuMat_8UC4.size(), cuMat_8UC4.type())
cuMat_8UC4_out = cuMat_8UC4.copyTo(cuMat_8UC4_dst)
self.assertTrue(cuMat_8UC4_out.cudaPtr() == cuMat_8UC4_dst.cudaPtr())
npMat_8UC4_out = cuMat_8UC4_out.download()
self.assertTrue(np.array_equal(npMat_8UC4, npMat_8UC4_out))
# out
cuMat_8UC4_out = cuMat_8UC4.copyTo()
npMat_8UC4_out = cuMat_8UC4_out.download()
self.assertTrue(np.array_equal(npMat_8UC4, npMat_8UC4_out))
# async
stream = cv.cuda.Stream()
# in/out
cuMat_8UC4 = cv.cuda_GpuMat(npMat_8UC4)
cuMat_8UC4_dst = cv.cuda_GpuMat(cuMat_8UC4.size(), cuMat_8UC4.type())
cuMat_8UC4_out = cuMat_8UC4.copyTo(cuMat_8UC4_dst, stream)
self.assertTrue(cuMat_8UC4_out.cudaPtr() == cuMat_8UC4_out.cudaPtr())
npMat_8UC4_out = cuMat_8UC4_dst.download(stream)
stream.waitForCompletion()
self.assertTrue(np.array_equal(npMat_8UC4, npMat_8UC4_out))
# out
cuMat_8UC4_out = cuMat_8UC4.copyTo(stream)
npMat_8UC4_out = cuMat_8UC4_out.download(stream)
stream.waitForCompletion()
self.assertTrue(np.array_equal(npMat_8UC4, npMat_8UC4_out))
def test_cuda_denoising(self):
self.assertEqual(True, hasattr(cv.cuda, 'fastNlMeansDenoising'))
self.assertEqual(True, hasattr(cv.cuda, 'fastNlMeansDenoisingColored'))
self.assertEqual(True, hasattr(cv.cuda, 'nonLocalMeans'))
if __name__ == '__main__':
NewOpenCVTests.bootstrap()
|
