Chiebot-Mirror
/
opencv
8
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Merge remote-tracking branch 'upstream/3.4' into merge-3.4

Browse Source
pull/21186/head
Alexander Alekhin 3 years ago
parent c3910807c5 35ff9af6ce
commit 8b4fa2605e
89 changed files with 1115 additions and 541 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 8
      modules/calib3d/src/calibration.cpp
  2. 6
      modules/calib3d/src/compat_ptsetreg.cpp
  3. 2
      modules/calib3d/test/test_cameracalibration.cpp
  4. 8
      modules/calib3d/test/test_fundam.cpp
  5. 34
      modules/calib3d/test/test_stereomatching.cpp
  6. 6
      modules/calib3d/test/test_undistort.cpp
  7. 20
      modules/core/src/array.cpp
  8. 60
      modules/core/src/datastructs.cpp
  9. 26
      modules/core/src/dxt.cpp
  10. 4
      modules/core/src/lapack.cpp
  11. 4
      modules/core/src/matmul.simd.hpp
  12. 2
      modules/core/src/matrix_sparse.cpp
  13. 2
      modules/core/src/opencl/runtime/opencl_clblas.cpp
  14. 2
      modules/core/src/opencl/runtime/opencl_clfft.cpp
  15. 2
      modules/core/src/opencl/runtime/opencl_core.cpp
  16. 2
      modules/core/src/persistence.cpp
  17. 4
      modules/core/src/persistence_xml.cpp
  18. 2
      modules/core/src/persistence_yml.cpp
  19. 1
      modules/core/src/precomp.hpp
  20. 46
      modules/core/test/test_ds.cpp
  21. 2
      modules/core/test/test_dxt.cpp
  22. 6
      modules/core/test/test_mat.cpp
  23. 14
      modules/core/test/test_math.cpp
  24. 2
      modules/dnn/include/opencv2/dnn/all_layers.hpp
  25. 90
      modules/dnn/perf/perf_recurrent.cpp
  26. 113
      modules/dnn/src/dnn.cpp
  27. 18
      modules/dnn/src/layers/elementwise_layers.cpp
  28. 6
      modules/dnn/src/layers/fully_connected_layer.cpp
  29. 67
      modules/dnn/src/layers/layers_common.simd.hpp
  30. 2
      modules/dnn/src/layers/lrn_layer.cpp
  31. 103
      modules/dnn/src/layers/recurrent_layers.cpp
  32. 1
      modules/dnn/src/ocl4dnn/src/ocl4dnn_conv_spatial.cpp
  33. 70
      modules/dnn/src/onnx/onnx_graph_simplifier.cpp
  34. 42
      modules/dnn/src/onnx/onnx_importer.cpp
  35. 5
      modules/dnn/src/opencl/activations.cl
  36. 2
      modules/dnn/test/test_backends.cpp
  37. 24
      modules/dnn/test/test_caffe_importer.cpp
  38. 70
      modules/dnn/test/test_darknet_importer.cpp
  39. 25
      modules/dnn/test/test_halide_layers.cpp
  40. 68
      modules/dnn/test/test_layers.cpp
  41. 63
      modules/dnn/test/test_misc.cpp
  42. 210
      modules/dnn/test/test_onnx_importer.cpp
  43. 103
      modules/dnn/test/test_tf_importer.cpp
  44. 11
      modules/dnn/test/test_torch_importer.cpp
  45. 2
      modules/features2d/test/test_nearestneighbors.cpp
  46. 4
      modules/flann/include/opencv2/flann/nn_index.h
  47. 1
      modules/highgui/src/window_winrt.cpp
  48. 22
      modules/imgproc/src/approx.cpp
  49. 2
      modules/imgproc/src/bilateral_filter.simd.hpp
  50. 8
      modules/imgproc/src/color_lab.cpp
  51. 4
      modules/imgproc/src/contours.cpp
  52. 2
      modules/imgproc/src/convhull.cpp
  53. 2
      modules/imgproc/src/distransform.cpp
  54. 10
      modules/imgproc/src/drawing.cpp
  55. 6
      modules/imgproc/src/emd.cpp
  56. 8
      modules/imgproc/src/hough.cpp
  57. 2
      modules/imgproc/src/median_blur.simd.hpp
  58. 2
      modules/imgproc/src/moments.cpp
  59. 1
      modules/imgproc/src/precomp.hpp
  60. 8
      modules/imgproc/src/resize.cpp
  61. 2
      modules/imgproc/src/samplers.cpp
  62. 6
      modules/imgproc/src/segmentation.cpp
  63. 2
      modules/imgproc/src/subdivision2d.cpp
  64. 4
      modules/imgproc/test/test_approxpoly.cpp
  65. 2
      modules/imgproc/test/test_bilateral_filter.cpp
  66. 2
      modules/imgproc/test/test_canny.cpp
  67. 41
      modules/imgproc/test/test_color.cpp
  68. 2
      modules/imgproc/test/test_contours.cpp
  69. 16
      modules/imgproc/test/test_convhull.cpp
  70. 2
      modules/imgproc/test/test_distancetransform.cpp
  71. 8
      modules/imgproc/test/test_filter.cpp
  72. 6
      modules/imgproc/test/test_imgwarp.cpp
  73. 2
      modules/imgproc/test/test_imgwarp_strict.cpp
  74. 2
      modules/imgproc/test/test_templmatch.cpp
  75. 2
      modules/imgproc/test/test_thresh.cpp
  76. 1
      modules/ml/src/precomp.hpp
  77. 2
      modules/ml/src/tree.cpp
  78. 8
      modules/objdetect/src/cascadedetect.cpp
  79. 4
      modules/objdetect/src/hog.cpp
  80. 47
      modules/objdetect/src/qrcode_encoder.cpp
  81. 12
      modules/objdetect/test/test_cascadeandhog.cpp
  82. 2
      modules/ts/src/ocl_perf.cpp
  83. 2
      modules/ts/src/ts.cpp
  84. 6
      modules/ts/src/ts_arrtest.cpp
  85. 8
      modules/ts/src/ts_func.cpp
  86. 8
      modules/ts/src/ts_perf.cpp
  87. 9
      modules/videoio/src/cap_ffmpeg_impl.hpp
  88. 5
      modules/videoio/src/cap_v4l.cpp
  89. 1
      modules/videoio/src/precomp.hpp

8
modules/calib3d/src/calibration.cpp
Unescape View File

@ -2329,7 +2329,7 @@ static double cvStereoCalibrateImpl( const CvMat* _objectPoints, const CvMat* _i
if( solver.state == CvLevMarq::CALC_J )
{
int iofs = (nimages+1)*6 + k*NINTRINSIC, eofs = (i+1)*6;
assert( JtJ && JtErr );
CV_Assert( JtJ && JtErr );
Mat _JtJ(cvarrToMat(JtJ)), _JtErr(cvarrToMat(JtErr));
@ -3140,7 +3140,7 @@ cvRQDecomp3x3( const CvMat *matrixM, CvMat *matrixR, CvMat *matrixQ,
CvMat Qx = cvMat(3, 3, CV_64F, _Qx);
cvMatMul(&M, &Qx, &R);
assert(fabs(matR[2][1]) < FLT_EPSILON);
CV_DbgAssert(fabs(matR[2][1]) < FLT_EPSILON);
matR[2][1] = 0;
/* Find Givens rotation for y axis. */
@ -3159,7 +3159,7 @@ cvRQDecomp3x3( const CvMat *matrixM, CvMat *matrixR, CvMat *matrixQ,
CvMat Qy = cvMat(3, 3, CV_64F, _Qy);
cvMatMul(&R, &Qy, &M);
assert(fabs(matM[2][0]) < FLT_EPSILON);
CV_DbgAssert(fabs(matM[2][0]) < FLT_EPSILON);
matM[2][0] = 0;
/* Find Givens rotation for z axis. */
@ -3179,7 +3179,7 @@ cvRQDecomp3x3( const CvMat *matrixM, CvMat *matrixR, CvMat *matrixQ,
CvMat Qz = cvMat(3, 3, CV_64F, _Qz);
cvMatMul(&M, &Qz, &R);
assert(fabs(matR[1][0]) < FLT_EPSILON);
CV_DbgAssert(fabs(matR[1][0]) < FLT_EPSILON);
matR[1][0] = 0;
// Solve the decomposition ambiguity.

6
modules/calib3d/src/compat_ptsetreg.cpp
Unescape View File

@ -122,7 +122,7 @@ bool CvLevMarq::update( const CvMat*& _param, CvMat*& matJ, CvMat*& _err )
{
matJ = _err = 0;
assert( !err.empty() );
CV_Assert( !err.empty() );
if( state == DONE )
{
_param = param;
@ -155,7 +155,7 @@ bool CvLevMarq::update( const CvMat*& _param, CvMat*& matJ, CvMat*& _err )
return true;
}
assert( state == CHECK_ERR );
CV_Assert( state == CHECK_ERR );
errNorm = cvNorm( err, 0, CV_L2 );
if( errNorm > prevErrNorm )
{
@ -223,7 +223,7 @@ bool CvLevMarq::updateAlt( const CvMat*& _param, CvMat*& _JtJ, CvMat*& _JtErr, d
return true;
}
assert( state == CHECK_ERR );
CV_Assert( state == CHECK_ERR );
if( errNorm > prevErrNorm )
{
if( ++lambdaLg10 <= 16 )

2
modules/calib3d/test/test_cameracalibration.cpp
Unescape View File

@ -1072,7 +1072,7 @@ void CV_ProjectPointsTest::run(int)
imgPoints, dpdrot, dpdt, dpdf, dpdc, dpddist, 0 );
// calculate and check image points
assert( (int)imgPoints.size() == pointCount );
CV_Assert( (int)imgPoints.size() == pointCount );
vector<Point2f>::const_iterator it = imgPoints.begin();
for( int i = 0; i < pointCount; i++, ++it )
{

8
modules/calib3d/test/test_fundam.cpp
Unescape View File

@ -57,7 +57,7 @@ static int cvTsRodrigues( const CvMat* src, CvMat* dst, CvMat* jacobian )
if( jacobian )
{
assert( (jacobian->rows == 9 && jacobian->cols == 3) ||
CV_Assert( (jacobian->rows == 9 && jacobian->cols == 3) ||
(jacobian->rows == 3 && jacobian->cols == 9) );
}
@ -66,7 +66,7 @@ static int cvTsRodrigues( const CvMat* src, CvMat* dst, CvMat* jacobian )
double r[3], theta;
CvMat _r = cvMat( src->rows, src->cols, CV_MAKETYPE(CV_64F,CV_MAT_CN(src->type)), r);
assert( dst->rows == 3 && dst->cols == 3 );
CV_Assert( dst->rows == 3 && dst->cols == 3 );
cvConvert( src, &_r );
@ -321,7 +321,7 @@ static int cvTsRodrigues( const CvMat* src, CvMat* dst, CvMat* jacobian )
}
else
{
assert(0);
CV_Assert(0);
return 0;
}
@ -407,7 +407,7 @@ static void test_convertHomogeneous( const Mat& _src, Mat& _dst )
}
else
{
assert( count == dst.cols );
CV_Assert( count == dst.cols );
ddims = dst.channels()*dst.rows;
if( dst.rows == 1 )
{

34
modules/calib3d/test/test_stereomatching.cpp
Unescape View File

@ -406,7 +406,7 @@ void CV_StereoMatchingTest::run(int)
{
string dataPath = ts->get_data_path() + "cv/";
string algorithmName = name;
assert( !algorithmName.empty() );
CV_Assert( !algorithmName.empty() );
if( dataPath.empty() )
{
ts->printf( cvtest::TS::LOG, "dataPath is empty" );
@ -553,22 +553,22 @@ int CV_StereoMatchingTest::processStereoMatchingResults( FileStorage& fs, int ca
{
// rightDisp is not used in current test virsion
int code = cvtest::TS::OK;
assert( fs.isOpened() );
assert( trueLeftDisp.type() == CV_32FC1 );
assert( trueRightDisp.empty() || trueRightDisp.type() == CV_32FC1 );
assert( leftDisp.type() == CV_32FC1 && (rightDisp.empty() || rightDisp.type() == CV_32FC1) );
CV_Assert( fs.isOpened() );
CV_Assert( trueLeftDisp.type() == CV_32FC1 );
CV_Assert( trueRightDisp.empty() || trueRightDisp.type() == CV_32FC1 );
CV_Assert( leftDisp.type() == CV_32FC1 && (rightDisp.empty() || rightDisp.type() == CV_32FC1) );
// get masks for unknown ground truth disparity values
Mat leftUnknMask, rightUnknMask;
DatasetParams params = datasetsParams[caseDatasets[caseIdx]];
absdiff( trueLeftDisp, Scalar(params.dispUnknVal), leftUnknMask );
leftUnknMask = leftUnknMask < std::numeric_limits<float>::epsilon();
assert(leftUnknMask.type() == CV_8UC1);
CV_Assert(leftUnknMask.type() == CV_8UC1);
if( !trueRightDisp.empty() )
{
absdiff( trueRightDisp, Scalar(params.dispUnknVal), rightUnknMask );
rightUnknMask = rightUnknMask < std::numeric_limits<float>::epsilon();
assert(rightUnknMask.type() == CV_8UC1);
CV_Assert(rightUnknMask.type() == CV_8UC1);
}
// calculate errors
@ -623,7 +623,7 @@ int CV_StereoMatchingTest::readDatasetsParams( FileStorage& fs )
}
datasetsParams.clear();
FileNode fn = fs.getFirstTopLevelNode();
assert(fn.isSeq());
CV_Assert(fn.isSeq());
for( int i = 0; i < (int)fn.size(); i+=3 )
{
String _name = fn[i];
@ -649,7 +649,7 @@ int CV_StereoMatchingTest::readRunParams( FileStorage& fs )
void CV_StereoMatchingTest::writeErrors( const string& errName, const vector<float>& errors, FileStorage* fs )
{
assert( (int)errors.size() == ERROR_KINDS_COUNT );
CV_Assert( (int)errors.size() == ERROR_KINDS_COUNT );
vector<float>::const_iterator it = errors.begin();
if( fs )
for( int i = 0; i < ERROR_KINDS_COUNT; i++, ++it )
@ -696,9 +696,9 @@ void CV_StereoMatchingTest::readROI( FileNode& fn, Rect& validROI )
int CV_StereoMatchingTest::compareErrors( const vector<float>& calcErrors, const vector<float>& validErrors,
const vector<float>& eps, const string& errName )
{
assert( (int)calcErrors.size() == ERROR_KINDS_COUNT );
assert( (int)validErrors.size() == ERROR_KINDS_COUNT );
assert( (int)eps.size() == ERROR_KINDS_COUNT );
CV_Assert( (int)calcErrors.size() == ERROR_KINDS_COUNT );
CV_Assert( (int)validErrors.size() == ERROR_KINDS_COUNT );
CV_Assert( (int)eps.size() == ERROR_KINDS_COUNT );
vector<float>::const_iterator calcIt = calcErrors.begin(),
validIt = validErrors.begin(),
epsIt = eps.begin();
@ -757,7 +757,7 @@ protected:
{
int code = CV_StereoMatchingTest::readRunParams( fs );
FileNode fn = fs.getFirstTopLevelNode();
assert(fn.isSeq());
CV_Assert(fn.isSeq());
for( int i = 0; i < (int)fn.size(); i+=5 )
{
String caseName = fn[i], datasetName = fn[i+1];
@ -776,8 +776,8 @@ protected:
Rect& calcROI, Mat& leftDisp, Mat& /*rightDisp*/, int caseIdx )
{
RunParams params = caseRunParams[caseIdx];
assert( params.ndisp%16 == 0 );
assert( _leftImg.type() == CV_8UC3 && _rightImg.type() == CV_8UC3 );
CV_Assert( params.ndisp%16 == 0 );
CV_Assert( _leftImg.type() == CV_8UC3 && _rightImg.type() == CV_8UC3 );
Mat leftImg; cvtColor( _leftImg, leftImg, COLOR_BGR2GRAY );
Mat rightImg; cvtColor( _rightImg, rightImg, COLOR_BGR2GRAY );
@ -883,7 +883,7 @@ protected:
{
int code = CV_StereoMatchingTest::readRunParams(fs);
FileNode fn = fs.getFirstTopLevelNode();
assert(fn.isSeq());
CV_Assert(fn.isSeq());
for( int i = 0; i < (int)fn.size(); i+=5 )
{
String caseName = fn[i], datasetName = fn[i+1];
@ -902,7 +902,7 @@ protected:
Rect& calcROI, Mat& leftDisp, Mat& /*rightDisp*/, int caseIdx )
{
RunParams params = caseRunParams[caseIdx];
assert( params.ndisp%16 == 0 );
CV_Assert( params.ndisp%16 == 0 );
Ptr<StereoSGBM> sgbm = StereoSGBM::create( 0, params.ndisp, params.winSize,
10*params.winSize*params.winSize,
40*params.winSize*params.winSize,

6
modules/calib3d/test/test_undistort.cpp
Unescape View File

@ -1108,7 +1108,7 @@ static void test_remap( const Mat& src, Mat& dst, const Mat& mapx, const Mat& ma
}
break;
default:
assert(0);
CV_Assert(0);
}
}
}
@ -1240,7 +1240,7 @@ int CV_ImgWarpBaseTest::prepare_test_case( int test_case_idx )
}
break;
default:
assert(0);
CV_Assert(0);
}
/*switch( depth )
@ -1258,7 +1258,7 @@ int CV_ImgWarpBaseTest::prepare_test_case( int test_case_idx )
((float*)ptr)[j] = (float)buffer[j];
break;
default:
assert(0);
CV_Assert(0);
}*/
cv::Mat src(1, cols*cn, CV_32F, &buffer[0]);
cv::Mat dst(1, cols*cn, depth, ptr);

20
modules/core/src/array.cpp
Unescape View File

@ -497,7 +497,7 @@ cvInitNArrayIterator( int count, CvArr** arrs,
// returns zero value if iteration is finished, non-zero otherwise
CV_IMPL int cvNextNArraySlice( CvNArrayIterator* iterator )
{
assert( iterator != 0 );
CV_Assert( iterator != 0 );
int i, dims;
for( dims = iterator->dims; dims > 0; dims-- )
@ -648,7 +648,7 @@ icvGetNodePtr( CvSparseMat* mat, const int* idx, int* _type,
int i, tabidx;
unsigned hashval = 0;
CvSparseNode *node;
assert( CV_IS_SPARSE_MAT( mat ));
CV_Assert( CV_IS_SPARSE_MAT( mat ));
if( !precalc_hashval )
{
@ -697,7 +697,7 @@ icvGetNodePtr( CvSparseMat* mat, const int* idx, int* _type,
int newrawsize = newsize*sizeof(newtable[0]);
CvSparseMatIterator iterator;
assert( (newsize & (newsize - 1)) == 0 );
CV_Assert( (newsize & (newsize - 1)) == 0 );
// resize hash table
newtable = (void**)cvAlloc( newrawsize );
@ -742,7 +742,7 @@ icvDeleteNode( CvSparseMat* mat, const int* idx, unsigned* precalc_hashval )
int i, tabidx;
unsigned hashval = 0;
CvSparseNode *node, *prev = 0;
assert( CV_IS_SPARSE_MAT( mat ));
CV_Assert( CV_IS_SPARSE_MAT( mat ));
if( !precalc_hashval )
{
@ -1462,7 +1462,7 @@ cvScalarToRawData( const CvScalar* scalar, void* data, int type, int extend_to_1
int cn = CV_MAT_CN( type );
int depth = type & CV_MAT_DEPTH_MASK;
assert( scalar && data );
CV_Assert( scalar && data );
if( (unsigned)(cn - 1) >= 4 )
CV_Error( CV_StsOutOfRange, "The number of channels must be 1, 2, 3 or 4" );
@ -1509,7 +1509,7 @@ cvScalarToRawData( const CvScalar* scalar, void* data, int type, int extend_to_1
((double*)data)[cn] = (double)(scalar->val[cn]);
break;
default:
assert(0);
CV_Assert(0);
CV_Error( CV_BadDepth, "" );
}
@ -1534,7 +1534,7 @@ cvRawDataToScalar( const void* data, int flags, CvScalar* scalar )
{
int cn = CV_MAT_CN( flags );
assert( scalar && data );
CV_Assert( scalar && data );
if( (unsigned)(cn - 1) >= 4 )
CV_Error( CV_StsOutOfRange, "The number of channels must be 1, 2, 3 or 4" );
@ -1572,7 +1572,7 @@ cvRawDataToScalar( const void* data, int flags, CvScalar* scalar )
scalar->val[cn] = ((double*)data)[cn];
break;
default:
assert(0);
CV_Assert(0);
CV_Error( CV_BadDepth, "" );
}
}
@ -2623,7 +2623,7 @@ cvReshapeMatND( const CvArr* arr,
{
CvMatND* mat = (CvMatND*)arr;
assert( new_cn > 0 );
CV_Assert( new_cn > 0 );
int last_dim_size = mat->dim[mat->dims-1].size*CV_MAT_CN(mat->type);
int new_size = last_dim_size/new_cn;
@ -2901,7 +2901,7 @@ CV_IMPL IplImage *
cvCreateImage( CvSize size, int depth, int channels )
{
IplImage *img = cvCreateImageHeader( size, depth, channels );
assert( img );
CV_Assert( img );
cvCreateData( img );
return img;

60
modules/core/src/datastructs.cpp
Unescape View File

@ -97,7 +97,7 @@ icvInitMemStorage( CvMemStorage* storage, int block_size )
block_size = CV_STORAGE_BLOCK_SIZE;
block_size = cvAlign( block_size, CV_STRUCT_ALIGN );
assert( sizeof(CvMemBlock) % CV_STRUCT_ALIGN == 0 );
CV_Assert( sizeof(CvMemBlock) % CV_STRUCT_ALIGN == 0 );
memset( storage, 0, sizeof( *storage ));
storage->signature = CV_STORAGE_MAGIC_VAL;
@ -240,7 +240,7 @@ icvGoNextMemBlock( CvMemStorage * storage )
if( block == parent->top ) /* the single allocated block */
{
assert( parent->bottom == block );
CV_Assert( parent->bottom == block );
parent->top = parent->bottom = 0;
parent->free_space = 0;
}
@ -266,7 +266,7 @@ icvGoNextMemBlock( CvMemStorage * storage )
if( storage->top->next )
storage->top = storage->top->next;
storage->free_space = storage->block_size - sizeof(CvMemBlock);
assert( storage->free_space % CV_STRUCT_ALIGN == 0 );
CV_Assert( storage->free_space % CV_STRUCT_ALIGN == 0 );
}
@ -331,7 +331,7 @@ cvMemStorageAlloc( CvMemStorage* storage, size_t size )
if( size > INT_MAX )
CV_Error( CV_StsOutOfRange, "Too large memory block is requested" );
assert( storage->free_space % CV_STRUCT_ALIGN == 0 );
CV_Assert( storage->free_space % CV_STRUCT_ALIGN == 0 );
if( (size_t)storage->free_space < size )
{
@ -343,7 +343,7 @@ cvMemStorageAlloc( CvMemStorage* storage, size_t size )
}
ptr = ICV_FREE_PTR(storage);
assert( (size_t)ptr % CV_STRUCT_ALIGN == 0 );
CV_Assert( (size_t)ptr % CV_STRUCT_ALIGN == 0 );
storage->free_space = cvAlignLeft(storage->free_space - (int)size, CV_STRUCT_ALIGN );
return ptr;
@ -683,7 +683,7 @@ icvGrowSeq( CvSeq *seq, int in_front_of )
else
{
icvGoNextMemBlock( storage );
assert( storage->free_space >= delta );
CV_Assert( storage->free_space >= delta );
}
}
@ -716,7 +716,7 @@ icvGrowSeq( CvSeq *seq, int in_front_of )
* For used blocks it means current number
* of sequence elements in the block:
*/
assert( block->count % seq->elem_size == 0 && block->count > 0 );
CV_Assert( block->count % seq->elem_size == 0 && block->count > 0 );
if( !in_front_of )
{
@ -732,7 +732,7 @@ icvGrowSeq( CvSeq *seq, int in_front_of )
if( block != block->prev )
{
assert( seq->first->start_index == 0 );
CV_Assert( seq->first->start_index == 0 );
seq->first = block;
}
else
@ -760,7 +760,7 @@ icvFreeSeqBlock( CvSeq *seq, int in_front_of )
{
CvSeqBlock *block = seq->first;
assert( (in_front_of ? block : block->prev)->count == 0 );
CV_Assert( (in_front_of ? block : block->prev)->count == 0 );
if( block == block->prev ) /* single block case */
{
@ -775,7 +775,7 @@ icvFreeSeqBlock( CvSeq *seq, int in_front_of )
if( !in_front_of )
{
block = block->prev;
assert( seq->ptr == block->data );
CV_Assert( seq->ptr == block->data );
block->count = (int)(seq->block_max - seq->ptr);
seq->block_max = seq->ptr = block->prev->data +
@ -804,7 +804,7 @@ icvFreeSeqBlock( CvSeq *seq, int in_front_of )
block->next->prev = block->prev;
}
assert( block->count > 0 && block->count % seq->elem_size == 0 );
CV_Assert( block->count > 0 && block->count % seq->elem_size == 0 );
block->next = seq->free_blocks;
seq->free_blocks = block;
}
@ -861,7 +861,7 @@ cvFlushSeqWriter( CvSeqWriter * writer )
CvSeqBlock *block = first_block;
writer->block->count = (int)((writer->ptr - writer->block->data) / seq->elem_size);
assert( writer->block->count > 0 );
CV_Assert( writer->block->count > 0 );
do
{
@ -891,7 +891,7 @@ cvEndWriteSeq( CvSeqWriter * writer )
CvMemStorage *storage = seq->storage;
schar *storage_block_max = (schar *) storage->top + storage->block_size;
assert( writer->block->count > 0 );
CV_Assert( writer->block->count > 0 );
if( (unsigned)((storage_block_max - storage->free_space)
- seq->block_max) < CV_STRUCT_ALIGN )
@ -1147,7 +1147,7 @@ cvSeqPush( CvSeq *seq, const void *element )
icvGrowSeq( seq, 0 );
ptr = seq->ptr;
assert( ptr + elem_size <= seq->block_max /*&& ptr == seq->block_min */ );
CV_Assert( ptr + elem_size <= seq->block_max /*&& ptr == seq->block_min */ );
}
if( element )
@ -1183,7 +1183,7 @@ cvSeqPop( CvSeq *seq, void *element )
if( --(seq->first->prev->count) == 0 )
{
icvFreeSeqBlock( seq, 0 );
assert( seq->ptr == seq->block_max );
CV_Assert( seq->ptr == seq->block_max );
}
}
@ -1207,7 +1207,7 @@ cvSeqPushFront( CvSeq *seq, const void *element )
icvGrowSeq( seq, 1 );
block = seq->first;
assert( block->start_index > 0 );
CV_Assert( block->start_index > 0 );
}
ptr = block->data -= elem_size;
@ -1289,7 +1289,7 @@ cvSeqInsert( CvSeq *seq, int before_index, const void *element )
icvGrowSeq( seq, 0 );
ptr = seq->ptr + elem_size;
assert( ptr <= seq->block_max );
CV_Assert( ptr <= seq->block_max );
}
delta_index = seq->first->start_index;
@ -1307,7 +1307,7 @@ cvSeqInsert( CvSeq *seq, int before_index, const void *element )
block = prev_block;
/* Check that we don't fall into an infinite loop: */
assert( block != seq->first->prev );
CV_Assert( block != seq->first->prev );
}
before_index = (before_index - block->start_index + delta_index) * elem_size;
@ -1346,7 +1346,7 @@ cvSeqInsert( CvSeq *seq, int before_index, const void *element )
block = next_block;
/* Check that we don't fall into an infinite loop: */
assert( block != seq->first );
CV_Assert( block != seq->first );
}
before_index = (before_index - block->start_index + delta_index) * elem_size;
@ -1502,7 +1502,7 @@ cvSeqPushMulti( CvSeq *seq, const void *_elements, int count, int front )
icvGrowSeq( seq, 1 );
block = seq->first;
assert( block->start_index > 0 );
CV_Assert( block->start_index > 0 );
}
delta = MIN( block->start_index, count );
@ -1543,7 +1543,7 @@ cvSeqPopMulti( CvSeq *seq, void *_elements, int count, int front )
int delta = seq->first->prev->count;
delta = MIN( delta, count );
assert( delta > 0 );
CV_Assert( delta > 0 );
seq->first->prev->count -= delta;
seq->total -= delta;
@ -1568,7 +1568,7 @@ cvSeqPopMulti( CvSeq *seq, void *_elements, int count, int front )
int delta = seq->first->count;
delta = MIN( delta, count );
assert( delta > 0 );
CV_Assert( delta > 0 );
seq->first->count -= delta;
seq->total -= delta;
@ -2418,7 +2418,7 @@ cvSeqPartition( const CvSeq* seq, CvMemStorage* storage, CvSeq** labels,
root2->rank += root->rank == root2->rank;
root = root2;
}
assert( root->parent == 0 );
CV_Assert( root->parent == 0 );
// Compress path from node2 to the root:
while( node2->parent )
@ -2521,7 +2521,7 @@ cvSetAdd( CvSet* set, CvSetElem* element, CvSetElem** inserted_element )
((CvSetElem*)ptr)->flags = count | CV_SET_ELEM_FREE_FLAG;
((CvSetElem*)ptr)->next_free = (CvSetElem*)(ptr + elem_size);
}
assert( count <= CV_SET_ELEM_IDX_MASK+1 );
CV_Assert( count <= CV_SET_ELEM_IDX_MASK+1 );
((CvSetElem*)(ptr - elem_size))->next_free = 0;
set->first->prev->count += count - set->total;
set->total = count;
@ -2720,7 +2720,7 @@ cvFindGraphEdgeByPtr( const CvGraph* graph,
for( ; edge; edge = edge->next[ofs] )
{
ofs = start_vtx == edge->vtx[1];
assert( ofs == 1 || start_vtx == edge->vtx[0] );
CV_Assert( ofs == 1 || start_vtx == edge->vtx[0] );
if( edge->vtx[1] == end_vtx )
break;
}
@ -2784,7 +2784,7 @@ cvGraphAddEdgeByPtr( CvGraph* graph,
"vertex pointers coincide (or set to NULL)" );
edge = (CvGraphEdge*)cvSetNew( (CvSet*)(graph->edges) );
assert( edge->flags >= 0 );
CV_Assert( edge->flags >= 0 );
edge->vtx[0] = start_vtx;
edge->vtx[1] = end_vtx;
@ -2861,7 +2861,7 @@ cvGraphRemoveEdgeByPtr( CvGraph* graph, CvGraphVtx* start_vtx, CvGraphVtx* end_v
prev_ofs = ofs, prev_edge = edge, edge = edge->next[ofs] )
{
ofs = start_vtx == edge->vtx[1];
assert( ofs == 1 || start_vtx == edge->vtx[0] );
CV_Assert( ofs == 1 || start_vtx == edge->vtx[0] );
if( edge->vtx[1] == end_vtx )
break;
}
@ -2879,7 +2879,7 @@ cvGraphRemoveEdgeByPtr( CvGraph* graph, CvGraphVtx* start_vtx, CvGraphVtx* end_v
prev_ofs = ofs, prev_edge = edge, edge = edge->next[ofs] )
{
ofs = end_vtx == edge->vtx[1];
assert( ofs == 1 || end_vtx == edge->vtx[0] );
CV_Assert( ofs == 1 || end_vtx == edge->vtx[0] );
if( edge->vtx[0] == start_vtx )
break;
}
@ -3396,7 +3396,7 @@ cvInsertNodeIntoTree( void* _node, void* _parent, void* _frame )
node->v_prev = _parent != _frame ? parent : 0;
node->h_next = parent->v_next;
assert( parent->v_next != node );
CV_Assert( parent->v_next != node );
if( parent->v_next )
parent->v_next->h_prev = node;
@ -3430,7 +3430,7 @@ cvRemoveNodeFromTree( void* _node, void* _frame )
if( parent )
{
assert( parent->v_next == node );
CV_Assert( parent->v_next == node );
parent->v_next = node->h_next;
}
}

26
modules/core/src/dxt.cpp
Unescape View File

@ -238,7 +238,7 @@ DFTInit( int n0, int nf, const int* factors, int* itab, int elem_size, void* _wa
else
{
// radix[] is initialized from index 'nf' down to zero
assert (nf < 34);
CV_Assert (nf < 34);
radix[nf] = 1;
digits[nf] = 0;
for( i = 0; i < nf; i++ )
@ -374,7 +374,7 @@ DFTInit( int n0, int nf, const int* factors, int* itab, int elem_size, void* _wa
else
{
Complex<float>* wave = (Complex<float>*)_wave;
assert( elem_size == sizeof(Complex<float>) );
CV_Assert( elem_size == sizeof(Complex<float>) );
wave[0].re = 1.f;
wave[0].im = 0.f;
@ -874,13 +874,13 @@ DFT(const OcvDftOptions & c, const Complex<T>* src, Complex<T>* dst)
// 0. shuffle data
if( dst != src )
{
assert( !c.noPermute );
CV_Assert( !c.noPermute );
if( !inv )
{
for( i = 0; i <= n - 2; i += 2, itab += 2*tab_step )
{
int k0 = itab[0], k1 = itab[tab_step];
assert( (unsigned)k0 < (unsigned)n && (unsigned)k1 < (unsigned)n );
CV_Assert( (unsigned)k0 < (unsigned)n && (unsigned)k1 < (unsigned)n );
dst[i] = src[k0]; dst[i+1] = src[k1];
}
@ -892,7 +892,7 @@ DFT(const OcvDftOptions & c, const Complex<T>* src, Complex<T>* dst)
for( i = 0; i <= n - 2; i += 2, itab += 2*tab_step )
{
int k0 = itab[0], k1 = itab[tab_step];
assert( (unsigned)k0 < (unsigned)n && (unsigned)k1 < (unsigned)n );
CV_Assert( (unsigned)k0 < (unsigned)n && (unsigned)k1 < (unsigned)n );
t.re = src[k0].re; t.im = -src[k0].im;
dst[i] = t;
t.re = src[k1].re; t.im = -src[k1].im;
@ -921,7 +921,7 @@ DFT(const OcvDftOptions & c, const Complex<T>* src, Complex<T>* dst)
for( i = 0; i < n2; i += 2, itab += tab_step*2 )
{
j = itab[0];
assert( (unsigned)j < (unsigned)n2 );
CV_Assert( (unsigned)j < (unsigned)n2 );
CV_SWAP(dst[i+1], dsth[j], t);
if( j > i )
@ -938,7 +938,7 @@ DFT(const OcvDftOptions & c, const Complex<T>* src, Complex<T>* dst)
for( i = 0; i < n; i++, itab += tab_step )
{
j = itab[0];
assert( (unsigned)j < (unsigned)n );
CV_Assert( (unsigned)j < (unsigned)n );
if( j > i )
CV_SWAP(dst[i], dst[j], t);
}
@ -1218,7 +1218,7 @@ RealDFT(const OcvDftOptions & c, const T* src, T* dst)
setIppErrorStatus();
#endif
}
assert( c.tab_size == n );
CV_Assert( c.tab_size == n );
if( n == 1 )
{
@ -1338,11 +1338,11 @@ CCSIDFT(const OcvDftOptions & c, const T* src, T* dst)
T save_s1 = 0.;
T t0, t1, t2, t3, t;
assert( c.tab_size == n );
CV_Assert( c.tab_size == n );
if( complex_input )
{
assert( src != dst );
CV_Assert( src != dst );
save_s1 = src[1];
((T*)src)[1] = src[0];
src++;
@ -3177,7 +3177,7 @@ protected:
}
else
{
assert( !inv );
CV_Assert( !inv );
CopyColumn( dbuf0, complex_elem_size, dptr0,
dst_step, len, complex_elem_size );
if( even )
@ -3874,7 +3874,7 @@ DCTInit( int n, int elem_size, void* _wave, int inv )
if( n == 1 )
return;
assert( (n&1) == 0 );
CV_Assert( (n&1) == 0 );
if( (n & (n - 1)) == 0 )
{
@ -3912,7 +3912,7 @@ DCTInit( int n, int elem_size, void* _wave, int inv )
else
{
Complex<float>* wave = (Complex<float>*)_wave;
assert( elem_size == sizeof(Complex<float>) );
CV_Assert( elem_size == sizeof(Complex<float>) );
w.re = (float)scale;
w.im = 0.f;

4
modules/core/src/lapack.cpp
Unescape View File

@ -1020,7 +1020,7 @@ double invert( InputArray _src, OutputArray _dst, int method )
}
else
{
assert( n == 1 );
CV_Assert( n == 1 );
if( type == CV_32FC1 )
{
@ -1208,7 +1208,7 @@ bool solve( InputArray _src, InputArray _src2arg, OutputArray _dst, int method )
}
else
{
assert( src.rows == 1 );
CV_Assert( src.rows == 1 );
if( type == CV_32FC1 )
{

4
modules/core/src/matmul.simd.hpp
Unescape View File

@ -169,7 +169,7 @@ GEMM_TransposeBlock( const uchar* src, size_t src_step,
}
break;
default:
assert(0);
CV_Assert(0);
return;
}
}
@ -2062,7 +2062,7 @@ MulTransposedR(const Mat& srcmat, const Mat& dstmat, const Mat& deltamat, double
if( delta && delta_cols < size.width )
{
assert( delta_cols == 1 );
CV_Assert( delta_cols == 1 );
buf_size *= 5;
}
buf.allocate(buf_size);

2
modules/core/src/matrix_sparse.cpp
Unescape View File

@ -638,7 +638,7 @@ void SparseMat::resizeHashTab(size_t newsize)
uchar* SparseMat::newNode(const int* idx, size_t hashval)
{
const int HASH_MAX_FILL_FACTOR=3;
assert(hdr);
CV_Assert(hdr);
size_t hsize = hdr->hashtab.size();
if( ++hdr->nodeCount > hsize*HASH_MAX_FILL_FACTOR )
{

2
modules/core/src/opencl/runtime/opencl_clblas.cpp
Unescape View File

@ -113,7 +113,7 @@ static void* openclamdblas_check_fn(int ID);
static void* openclamdblas_check_fn(int ID)
{
assert(ID >= 0 && ID < (int)(sizeof(openclamdblas_fn)/sizeof(openclamdblas_fn[0])));
CV_Assert(ID >= 0 && ID < (int)(sizeof(openclamdblas_fn)/sizeof(openclamdblas_fn[0])));
const struct DynamicFnEntry* e = openclamdblas_fn[ID];
void* func = CV_CL_GET_PROC_ADDRESS(e->fnName);
if (!func)

2
modules/core/src/opencl/runtime/opencl_clfft.cpp
Unescape View File

@ -113,7 +113,7 @@ static void* openclamdfft_check_fn(int ID);
static void* openclamdfft_check_fn(int ID)
{
assert(ID >= 0 && ID < (int)(sizeof(openclamdfft_fn)/sizeof(openclamdfft_fn[0])));
CV_Assert(ID >= 0 && ID < (int)(sizeof(openclamdfft_fn)/sizeof(openclamdfft_fn[0])));
const struct DynamicFnEntry* e = openclamdfft_fn[ID];
void* func = CV_CL_GET_PROC_ADDRESS(e->fnName);
if (!func)

2
modules/core/src/opencl/runtime/opencl_core.cpp
Unescape View File

@ -411,7 +411,7 @@ static void* opencl_gl_check_fn(int ID);
static void* opencl_gl_check_fn(int ID)
{
const struct DynamicFnEntry* e = NULL;
assert(ID >= 0 && ID < (int)(sizeof(opencl_gl_fn_list)/sizeof(opencl_gl_fn_list[0])));
CV_Assert(ID >= 0 && ID < (int)(sizeof(opencl_gl_fn_list)/sizeof(opencl_gl_fn_list[0])));
e = opencl_gl_fn_list[ID];
void* func = CV_CL_GET_PROC_ADDRESS(e->fnName);
if (!func)

2
modules/core/src/persistence.cpp
Unescape View File

@ -175,7 +175,7 @@ int decodeFormat( const char* dt, int* fmt_pairs, int max_len )
if( !dt || !len )
return 0;
assert( fmt_pairs != 0 && max_len > 0 );
CV_Assert( fmt_pairs != 0 && max_len > 0 );
fmt_pairs[0] = 0;
max_len *= 2;

4
modules/core/src/persistence_xml.cpp
Unescape View File

@ -387,7 +387,7 @@ public:
if( c == '-' )
{
assert( ptr[1] == '-' && ptr[2] == '>' );
CV_Assert( ptr[1] == '-' && ptr[2] == '>' );
mode = 0;
ptr += 3;
}
@ -694,7 +694,7 @@ public:
else if( *ptr == '!' )
{
tag_type = CV_XML_DIRECTIVE_TAG;
assert( ptr[1] != '-' || ptr[2] != '-' );
CV_Assert( ptr[1] != '-' || ptr[2] != '-' );
ptr++;
}
else

2
modules/core/src/persistence_yml.cpp
Unescape View File

@ -98,7 +98,7 @@ public:
/*
if( !FileNode::isFlow(parent_flags) )
fs->struct_indent -= CV_YML_INDENT + FileNode::isFlow(struct_flags);
assert( fs->struct_indent >= 0 );*/
CV_Assert( fs->struct_indent >= 0 );*/
}
void write(const char* key, int value)

1
modules/core/src/precomp.hpp
Unescape View File

@ -62,7 +62,6 @@
#include "opencv2/core/ocl.hpp"
#endif
#include <assert.h>
#include <ctype.h>
#include <float.h>
#include <limits.h>

46
modules/core/test/test_ds.cpp
Unescape View File

@ -33,7 +33,7 @@ static void cvTsReleaseSimpleSeq( CvTsSimpleSeq** seq )
static schar* cvTsSimpleSeqElem( CvTsSimpleSeq* seq, int index )
{
assert( 0 <= index && index < seq->count );
CV_Assert( 0 <= index && index < seq->count );
return seq->array + index * seq->elem_size;
}
@ -50,7 +50,11 @@ static void cvTsSimpleSeqShiftAndCopy( CvTsSimpleSeq* seq, int from_idx, int to_
if( from_idx == to_idx )
return;
assert( (from_idx > to_idx && !elem) || (from_idx < to_idx && elem) );
if (elem)
CV_Assert(from_idx < to_idx);
else
CV_Assert(from_idx > to_idx);
if( from_idx < seq->count )
{
@ -128,7 +132,7 @@ static void cvTsReleaseSimpleSet( CvTsSimpleSet** set_header )
static schar* cvTsSimpleSetFind( CvTsSimpleSet* set_header, int index )
{
int idx = index * set_header->elem_size;
assert( 0 <= index && index < set_header->max_count );
CV_Assert( 0 <= index && index < set_header->max_count );
return set_header->array[idx] ? set_header->array + idx + 1 : 0;
}
@ -136,11 +140,11 @@ static schar* cvTsSimpleSetFind( CvTsSimpleSet* set_header, int index )
static int cvTsSimpleSetAdd( CvTsSimpleSet* set_header, void* elem )
{
int idx, idx2;
assert( set_header->free_count > 0 );
CV_Assert( set_header->free_count > 0 );
idx = set_header->free_stack[--set_header->free_count];
idx2 = idx * set_header->elem_size;
assert( set_header->array[idx2] == 0 );
CV_Assert( set_header->array[idx2] == 0 );
set_header->array[idx2] = 1;
if( set_header->elem_size > 1 )
memcpy( set_header->array + idx2 + 1, elem, set_header->elem_size - 1 );
@ -152,9 +156,9 @@ static int cvTsSimpleSetAdd( CvTsSimpleSet* set_header, void* elem )
static void cvTsSimpleSetRemove( CvTsSimpleSet* set_header, int index )
{
assert( set_header->free_count < set_header->max_count &&
0 <= index && index < set_header->max_count );
assert( set_header->array[index * set_header->elem_size] == 1 );
CV_Assert( set_header->free_count < set_header->max_count &&
0 <= index && index < set_header->max_count );
CV_Assert( set_header->array[index * set_header->elem_size] == 1 );
set_header->free_stack[set_header->free_count++] = index;
set_header->array[index * set_header->elem_size] = 0;
@ -187,7 +191,7 @@ static CvTsSimpleGraph* cvTsCreateSimpleGraph( int max_vtx_count, int vtx_size,
{
CvTsSimpleGraph* graph;
assert( max_vtx_count > 1 && vtx_size >= 0 && edge_size >= 0 );
CV_Assert( max_vtx_count > 1 && vtx_size >= 0 && edge_size >= 0 );
graph = (CvTsSimpleGraph*)cvAlloc( sizeof(*graph) +
max_vtx_count * max_vtx_count * (edge_size + 1));
graph->vtx = cvTsCreateSimpleSet( max_vtx_count, vtx_size );
@ -235,13 +239,13 @@ static void cvTsSimpleGraphAddEdge( CvTsSimpleGraph* graph, int idx1, int idx2,
{
int i, t, n = graph->oriented ? 1 : 2;
assert( cvTsSimpleSetFind( graph->vtx, idx1 ) &&
cvTsSimpleSetFind( graph->vtx, idx2 ));
CV_Assert( cvTsSimpleSetFind( graph->vtx, idx1 ) &&
cvTsSimpleSetFind( graph->vtx, idx2 ));
for( i = 0; i < n; i++ )
{
int ofs = (idx1*graph->vtx->max_count + idx2)*graph->edge_size;
assert( graph->matrix[ofs] == 0 );
CV_Assert( graph->matrix[ofs] == 0 );
graph->matrix[ofs] = 1;
if( graph->edge_size > 1 )
memcpy( graph->matrix + ofs + 1, edge, graph->edge_size - 1 );
@ -255,13 +259,13 @@ static void cvTsSimpleGraphRemoveEdge( CvTsSimpleGraph* graph, int idx1, int id
{
int i, t, n = graph->oriented ? 1 : 2;
assert( cvTsSimpleSetFind( graph->vtx, idx1 ) &&
CV_Assert( cvTsSimpleSetFind( graph->vtx, idx1 ) &&
cvTsSimpleSetFind( graph->vtx, idx2 ));
for( i = 0; i < n; i++ )
{
int ofs = (idx1*graph->vtx->max_count + idx2)*graph->edge_size;
assert( graph->matrix[ofs] == 1 );
CV_Assert( graph->matrix[ofs] == 1 );
graph->matrix[ofs] = 0;
CV_SWAP( idx1, idx2, t );
}
@ -291,7 +295,7 @@ static int cvTsSimpleGraphVertexDegree( CvTsSimpleGraph* graph, int index )
int i, count = 0;
int edge_size = graph->edge_size;
int max_vtx_count = graph->vtx->max_count;
assert( cvTsSimpleGraphFindVertex( graph, index ) != 0 );
CV_Assert( cvTsSimpleGraphFindVertex( graph, index ) != 0 );
for( i = 0; i < max_vtx_count; i++ )
{
@ -301,7 +305,7 @@ static int cvTsSimpleGraphVertexDegree( CvTsSimpleGraph* graph, int index )
if( !graph->oriented )
{
assert( count % 2 == 0 );
CV_Assert( count % 2 == 0 );
count /= 2;
}
return count;
@ -609,7 +613,7 @@ int Core_SeqBaseTest::test_get_seq_elem( int _struct_idx, int iters )
CvTsSimpleSeq* sseq = (CvTsSimpleSeq*)simple_struct[_struct_idx];
struct_idx = _struct_idx;
assert( seq->total == sseq->count );
CV_Assert( seq->total == sseq->count );
if( sseq->count == 0 )
return 0;
@ -656,7 +660,7 @@ int Core_SeqBaseTest::test_get_seq_reading( int _struct_idx, int iters )
vector<schar> _elem(sseq->elem_size);
schar* elem = &_elem[0];
assert( total == sseq->count );
CV_Assert( total == sseq->count );
this->struct_idx = _struct_idx;
int pos = cvtest::randInt(rng) % 2;
@ -964,7 +968,7 @@ int Core_SeqBaseTest::test_seq_ops( int iters )
"The sequence doesn't become empty after clear" );
break;
default:
assert(0);
CV_Assert(0);
return -1;
}
@ -1903,7 +1907,7 @@ int Core_GraphScanTest::create_random_graph( int _struct_idx )
for( i = 0; i < vtx_count; i++ )
cvGraphAddVtx( graph );
assert( graph->active_count == vtx_count );
CV_Assert( graph->active_count == vtx_count );
for( i = 0; i < edge_count; i++ )
{
@ -1914,7 +1918,7 @@ int Core_GraphScanTest::create_random_graph( int _struct_idx )
cvGraphAddEdge( graph, j, k );
}
assert( graph->active_count == vtx_count && graph->edges->active_count <= edge_count );
CV_Assert( graph->active_count == vtx_count && graph->edges->active_count <= edge_count );
return 0;
}

2
modules/core/test/test_dxt.cpp
Unescape View File

@ -204,7 +204,7 @@ static void DCT_1D( const Mat& _src, Mat& _dst, int flags, const Mat& _wave=Mat(
}
}
else
assert(0);
CV_Assert(0);
}

6
modules/core/test/test_mat.cpp
Unescape View File

@ -28,7 +28,7 @@ protected:
template<class Type>
void testReduce( const Mat& src, Mat& sum, Mat& avg, Mat& max, Mat& min, int dim )
{
assert( src.channels() == 1 );
CV_Assert( src.channels() == 1 );
if( dim == 0 ) // row
{
sum.create( 1, src.cols, CV_64FC1 );
@ -138,7 +138,7 @@ int Core_ReduceTest::checkOp( const Mat& src, int dstType, int opType, const Mat
eps = 0.6;
}
assert( opRes.type() == CV_64FC1 );
CV_Assert( opRes.type() == CV_64FC1 );
Mat _dst, dst, diff;
cv::reduce( src, _dst, dim, opType, dstType );
_dst.convertTo( dst, CV_64FC1 );
@ -192,7 +192,7 @@ int Core_ReduceTest::checkCase( int srcType, int dstType, int dim, Size sz )
else if( srcType == CV_64FC1 )
testReduce<double>( src, sum, avg, max, min, dim );
else
assert( 0 );
CV_Assert( 0 );
// 1. sum
tempCode = checkOp( src, dstType, CV_REDUCE_SUM, sum, dim );

14
modules/core/test/test_math.cpp
Unescape View File

@ -1039,7 +1039,7 @@ static void cvTsPerspectiveTransform( const CvArr* _src, CvArr* _dst, const CvMa
}
else
{
assert( mat_depth == CV_64F );
CV_Assert( mat_depth == CV_64F );
for( i = 0; i < transmat->rows; i++ )
for( j = 0; j < cols; j++ )
mat[i*cols + j] = ((double*)(transmat->data.ptr + transmat->step*i))[j];
@ -1065,7 +1065,7 @@ static void cvTsPerspectiveTransform( const CvArr* _src, CvArr* _dst, const CvMa
buf[j] = ((double*)src)[j];
break;
default:
assert(0);
CV_Assert(0);
}
switch( cn )
@ -1095,7 +1095,7 @@ static void cvTsPerspectiveTransform( const CvArr* _src, CvArr* _dst, const CvMa
}
break;
default:
assert(0);
CV_Assert(0);
}
switch( depth )
@ -1109,7 +1109,7 @@ static void cvTsPerspectiveTransform( const CvArr* _src, CvArr* _dst, const CvMa
((double*)dst)[j] = buf[j];
break;
default:
assert(0);
CV_Assert(0);
}
}
}
@ -1458,8 +1458,8 @@ static double cvTsLU( CvMat* a, CvMat* b=NULL, CvMat* x=NULL, int* rank=0 )
double *a0 = a->data.db, *b0 = b ? b->data.db : 0;
double *x0 = x ? x->data.db : 0;
double t, det = 1.;
assert( CV_MAT_TYPE(a->type) == CV_64FC1 &&
(!b || CV_ARE_TYPES_EQ(a,b)) && (!x || CV_ARE_TYPES_EQ(a,x)));
CV_Assert( CV_MAT_TYPE(a->type) == CV_64FC1 &&
(!b || CV_ARE_TYPES_EQ(a,b)) && (!x || CV_ARE_TYPES_EQ(a,x)));
for( i = 0; i < Nm; i++ )
{
@ -1514,7 +1514,7 @@ static double cvTsLU( CvMat* a, CvMat* b=NULL, CvMat* x=NULL, int* rank=0 )
if( x )
{
assert( b );
CV_Assert( b );
for( i = N-1; i >= 0; i-- )
{

2
modules/dnn/include/opencv2/dnn/all_layers.hpp
Unescape View File

@ -545,6 +545,8 @@ CV__DNN_INLINE_NS_BEGIN
class CV_EXPORTS ELULayer : public ActivationLayer
{
public:
float alpha;
static Ptr<ELULayer> create(const LayerParams &params);
};

90
modules/dnn/perf/perf_recurrent.cpp
Unescape View File

@ -0,0 +1,90 @@
// This file is part of OpenCV project.
// It is subject to the license terms in the LICENSE file found in the top-level directory
// of this distribution and at http://opencv.org/license.html.
#include "perf_precomp.hpp"
namespace opencv_test {
struct LstmParams {
// Batch size
int nrSamples;
// Size of the input vector
int inputSize;
// Size of the internal state vector
int hiddenSize;
// Number of timesteps for the LSTM
int nrSteps;
};
static inline void PrintTo(const LstmParams& params, ::std::ostream* os) {
(*os) << "BATCH=" << params.nrSamples
<< ", IN=" << params.inputSize
<< ", HIDDEN=" << params.hiddenSize
<< ", TS=" << params.nrSteps;
}
static const LstmParams testLstmConfigs[] = {
{1, 192, 192, 100},
{1, 1024, 192, 100},
{1, 64, 192, 100},
{1, 192, 512, 100},
{64, 192, 192, 2},
{64, 1024, 192, 2},
{64, 64, 192, 2},
{64, 192, 512, 2},
{128, 192, 192, 2},
{128, 1024, 192, 2},
{128, 64, 192, 2},
{128, 192, 512, 2}
};
class Layer_LSTM : public TestBaseWithParam<LstmParams> {};
PERF_TEST_P_(Layer_LSTM, lstm) {
const LstmParams& params = GetParam();
LayerParams lp;
lp.type = "LSTM";
lp.name = "testLstm";
lp.set("produce_cell_output", false);
lp.set("use_timestamp_dim", true);
Mat weightH(params.hiddenSize * 4, params.hiddenSize, CV_32FC1, cv::Scalar(0));
Mat weightX(params.hiddenSize * 4, params.inputSize, CV_32FC1, cv::Scalar(0));
Mat bias(params.hiddenSize * 4, 1, CV_32FC1, cv::Scalar(0));
Mat hInternal(params.nrSteps, params.hiddenSize, CV_32FC1, cv::Scalar(0));
Mat cInternal(params.nrSteps, params.hiddenSize, CV_32FC1, cv::Scalar(0));
lp.blobs.push_back(weightH);
lp.blobs.push_back(weightX);
lp.blobs.push_back(bias);
lp.blobs.push_back(hInternal);
lp.blobs.push_back(cInternal);
std::vector<int> inputDims;
inputDims.push_back(params.nrSamples);
inputDims.push_back(params.nrSteps);
inputDims.push_back(params.inputSize);
Mat input(inputDims.size(), inputDims.data(), CV_32FC1);
input = cv::Scalar(0);
Net net;
net.addLayerToPrev(lp.name, lp.type, lp);
net.setInput(input);
// Warm up
std::vector<Mat> outputs(2);
net.forward(outputs, "testLstm");
TEST_CYCLE()
{
net.forward(outputs, "testLstm");
}
SANITY_CHECK_NOTHING();
}
INSTANTIATE_TEST_CASE_P(/**/, Layer_LSTM, testing::ValuesIn(testLstmConfigs));
} // namespace

113
modules/dnn/src/dnn.cpp
Unescape View File

@ -646,29 +646,26 @@ struct DataLayer : public Layer
CV_TRACE_FUNCTION();
CV_TRACE_ARG_VALUE(name, "name", name.c_str());
// FIXIT: add wrapper without exception suppression
CV_OCL_RUN(IS_DNN_OPENCL_TARGET(preferableTarget),
forward_ocl(inputs_arr, outputs_arr, internals_arr))
if (outputs_arr.depth() == CV_16S)
{
forward_fallback(inputs_arr, outputs_arr, internals_arr);
return;
}
bool isFP16 = outputs_arr.depth() == CV_16S;
std::vector<Mat> outputs, internals;
outputs_arr.getMatVector(outputs);
internals_arr.getMatVector(internals);
// Supported modes:
// | Input type | Output type |
// | fp32 | fp32 |
// | uint8 | fp32 |
for (int i = 0; i < inputsData.size(); ++i)
{
double scale = scaleFactors[i];
Scalar& mean = means[i];
CV_Assert(mean == Scalar() || inputsData[i].size[1] <= 4);
CV_CheckTypeEQ(outputs[i].type(), CV_32FC1, "");
if (isFP16)
CV_CheckTypeEQ(outputs[i].type(), CV_16SC1, "");
else
CV_CheckTypeEQ(outputs[i].type(), CV_32FC1, "");
bool singleMean = true;
for (int j = 1; j < std::min(4, inputsData[i].size[1]) && singleMean; ++j)
@ -678,34 +675,49 @@ struct DataLayer : public Layer
if (singleMean)
{
inputsData[i].convertTo(outputs[i], CV_32F, scale, -mean[0] * scale);
if (isFP16)
{
Mat input_f32;
inputsData[i].convertTo(input_f32, CV_32F, scale, -mean[0] * scale);
convertFp16(input_f32, outputs[i]);
}
else
{
inputsData[i].convertTo(outputs[i], CV_32F, scale, -mean[0] * scale);
}
}
else
{
for (int n = 0; n < inputsData[i].size[0]; ++n)
{
for (int c = 0; c < inputsData[i].size[1]; ++c)
{
Mat inp = getPlane(inputsData[i], n, c);
Mat out = getPlane(outputs[i], n, c);
inp.convertTo(out, CV_32F, scale, -mean[c] * scale);
if (isFP16)
{
Mat input_f32;
inp.convertTo(input_f32, CV_32F, scale, -mean[c] * scale);
convertFp16(input_f32, out);
}
else
{
inp.convertTo(out, CV_32F, scale, -mean[c] * scale);
}
}
}
}
}
}
#ifdef HAVE_OPENCL
std::vector<Mat> tmp_expressions;
bool forward_ocl(InputArrayOfArrays, OutputArrayOfArrays outputs_, OutputArrayOfArrays internals_)
{
// Supported modes:
// | Input type | Output type |
// | fp32 | fp32 |
// | fp32 | fp16 |
// | uint8 | fp32 |
bool isFP16 = outputs_.depth() == CV_16S;
std::vector<UMat> outputs;
outputs_.getUMatVector(outputs);
tmp_expressions.clear();
for (int i = 0; i < inputsData.size(); ++i)
{
Mat inputData = inputsData[i];
@ -713,58 +725,55 @@ struct DataLayer : public Layer
double scale = scaleFactors[i];
Scalar& mean = means[i];
CV_Assert(mean == Scalar() || inputsData[i].size[1] <= 4);
CV_Assert(mean == Scalar() || inputData.size[1] <= 4);
if (isFP16)
CV_CheckTypeEQ(outputs[i].type(), CV_16SC1, "");
else
CV_CheckTypeEQ(outputs[i].type(), CV_32FC1, "");
bool singleMean = true;
for (int j = 1; j < std::min(4, inputsData[i].size[1]) && singleMean; ++j)
for (int j = 1; j < std::min(4, inputData.size[1]) && singleMean; ++j)
{
singleMean = mean[j] == mean[j - 1];
}
if (outputs_.depth() == CV_16S)
if (singleMean)
{
if (singleMean)
if (isFP16)
{
tmp_expressions.push_back(Mat(scale * (inputsData[i] - mean[0])));
convertFp16(tmp_expressions.back(), outputs[i]);
UMat input_i;
inputData.convertTo(input_i, CV_32F, scale, -mean[0] * scale);
convertFp16(input_i, outputs[i]);
}
else
{
for (int n = 0; n < inputsData[i].size[0]; ++n)
for (int c = 0; c < inputsData[i].size[1]; ++c)
{
Mat inp = getPlane(inputsData[i], n, c);
std::vector<cv::Range> plane(4, Range::all());
plane[0] = Range(n, n + 1);
plane[1] = Range(c, c + 1);
UMat out = outputs[i](plane).reshape(1, inp.dims, inp.size);
tmp_expressions.push_back(scale * (inp - mean[c]));
convertFp16(tmp_expressions.back(), out);
}
inputData.convertTo(outputs[i], CV_32F, scale, -mean[0] * scale);
}
}
else
{
CV_Assert(outputs_.depth() == CV_32F);
if (singleMean)
{
inputsData[i].convertTo(outputs[i], CV_32F, scale, -mean[0] * scale);
}
else
for (int n = 0; n < inputData.size[0]; ++n)
{
for (int n = 0; n < inputsData[i].size[0]; ++n)
for (int c = 0; c < inputsData[i].size[1]; ++c)
{
Mat inp = getPlane(inputsData[i], n, c);
for (int c = 0; c < inputData.size[1]; ++c)
{
Mat inp = getPlane(inputData, n, c);
std::vector<cv::Range> plane(4, Range::all());
plane[0] = Range(n, n + 1);
plane[1] = Range(c, c + 1);
UMat out = outputs[i](plane).reshape(1, inp.dims, inp.size);
std::vector<cv::Range> plane(4, Range::all());
plane[0] = Range(n, n + 1);
plane[1] = Range(c, c + 1);
UMat out = outputs[i](plane).reshape(1, inp.dims, inp.size);
if (isFP16)
{
UMat input_i;
inp.convertTo(input_i, CV_32F, scale, -mean[c] * scale);
convertFp16(input_i, out);
}
else
{
inp.convertTo(out, CV_32F, scale, -mean[c] * scale);
}
}
}
}
}

18
modules/dnn/src/layers/elementwise_layers.cpp
Unescape View File

@ -987,6 +987,9 @@ const char* const SigmoidFunctor::BaseDefaultFunctor<SigmoidFunctor>::ocl_kernel
struct ELUFunctor : public BaseDefaultFunctor<ELUFunctor>
{
typedef ELULayer Layer;
float alpha;
explicit ELUFunctor(float alpha_ = 1.f) : alpha(alpha_) {}
bool supportBackend(int backendId, int)
{
@ -998,7 +1001,12 @@ struct ELUFunctor : public BaseDefaultFunctor<ELUFunctor>
inline float calculate(float x) const
{
return x >= 0.f ? x : exp(x) - 1.f;
return x >= 0.f ? x : alpha * (exp(x) - 1.f);
}
inline void setKernelParams(ocl::Kernel& kernel) const
{
kernel.set(3, alpha);
}
#ifdef HAVE_CUDA
@ -1012,7 +1020,7 @@ struct ELUFunctor : public BaseDefaultFunctor<ELUFunctor>
void attachHalide(const Halide::Expr& input, Halide::Func& top)
{
Halide::Var x("x"), y("y"), c("c"), n("n");
top(x, y, c, n) = select(input >= 0.0f, input, exp(input) - 1);
top(x, y, c, n) = select(input >= 0.0f, input, alpha * (exp(input) - 1));
}
#endif // HAVE_HALIDE
@ -1026,7 +1034,7 @@ struct ELUFunctor : public BaseDefaultFunctor<ELUFunctor>
#ifdef HAVE_DNN_NGRAPH
std::shared_ptr<ngraph::Node> initNgraphAPI(const std::shared_ptr<ngraph::Node>& node)
{
return std::make_shared<ngraph::op::Elu>(node, 1.0);
return std::make_shared<ngraph::op::Elu>(node, alpha);
}
#endif // HAVE_DNN_NGRAPH
@ -1856,8 +1864,10 @@ Ptr<SigmoidLayer> SigmoidLayer::create(const LayerParams& params)
Ptr<ELULayer> ELULayer::create(const LayerParams& params)
{
Ptr<ELULayer> l(new ElementWiseLayer<ELUFunctor>(ELUFunctor()));
float alpha = params.get<float>("alpha", 1.0f);
Ptr<ELULayer> l(new ElementWiseLayer<ELUFunctor>(ELUFunctor(alpha)));
l->setParamsFrom(params);
l->alpha = alpha;
return l;
}

6
modules/dnn/src/layers/fully_connected_layer.cpp
Unescape View File

@ -230,17 +230,17 @@ public:
#if CV_TRY_AVX512_SKX
if( useAVX512 )
opt_AVX512_SKX::fastGEMM1T( sptr, wptr, wstep, biasptr, dptr, nw, vecsize);
opt_AVX512_SKX::fastGEMM1T( sptr, wptr, wstep, biasptr, dptr, nw, vecsize_aligned);
else
#endif
#if CV_TRY_AVX2
if( useAVX2 )
opt_AVX2::fastGEMM1T( sptr, wptr, wstep, biasptr, dptr, nw, vecsize);
opt_AVX2::fastGEMM1T( sptr, wptr, wstep, biasptr, dptr, nw, vecsize_aligned);
else
#endif
#if CV_TRY_AVX
if( useAVX )
opt_AVX::fastGEMM1T( sptr, wptr, wstep, biasptr, dptr, nw, vecsize);
opt_AVX::fastGEMM1T( sptr, wptr, wstep, biasptr, dptr, nw, vecsize_aligned);
else
#endif
#if CV_TRY_RVV

67
modules/dnn/src/layers/layers_common.simd.hpp
Unescape View File

@ -550,13 +550,24 @@ void fastDepthwiseConv( const float* wptr,
_mm256_zeroupper();
}
// Used to generate the mask used when calculating tails
static const uint32_t tailMaskArray[15] = {
0, 0, 0, 0, 0, 0, 0, 0,
0xffffffffUL, 0xffffffffUL, 0xffffffffUL, 0xffffffffUL, 0xffffffffUL, 0xffffffffUL, 0xffffffffUL
};
// dst = vec * weights^t + bias
// Requires that vecsize is at least 8 or equal to 0 to avoid memory access problems. Does not require alignment.
void fastGEMM1T( const float* vec, const float* weights,
size_t wstep, const float* bias,
float* dst, int nvecs, int vecsize )
{
int i = 0;
CV_Assert(vecsize >= 8 || vecsize == 0);
__m256 tailMask = _mm256_loadu_ps(reinterpret_cast<const float*>(tailMaskArray) + (vecsize % 8));
for( ; i <= nvecs - 8; i += 8 )
{
const float* wptr = weights + i*wstep;
@ -565,18 +576,36 @@ void fastGEMM1T( const float* vec, const float* weights,
vs4 = _mm256_setzero_ps(), vs5 = _mm256_setzero_ps(),
vs6 = _mm256_setzero_ps(), vs7 = _mm256_setzero_ps();
for( int k = 0; k < vecsize; k += 8, wptr += 8 )
int k = 0;
for( ; k <= vecsize-8; k += 8, wptr += 8 )
{
__m256 v = _mm256_load_ps(vec + k);
vs0 = _mm256_fmadd_ps(_mm256_load_ps(wptr), v, vs0);
vs1 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep), v, vs1);
vs2 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep*2), v, vs2);
vs3 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep*3), v, vs3);
vs4 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep*4), v, vs4);
vs5 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep*5), v, vs5);
vs6 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep*6), v, vs6);
vs7 = _mm256_fmadd_ps(_mm256_load_ps(wptr + wstep*7), v, vs7);
__m256 v = _mm256_loadu_ps(vec + k);
vs0 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr), v, vs0);
vs1 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep), v, vs1);
vs2 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep*2), v, vs2);
vs3 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep*3), v, vs3);
vs4 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep*4), v, vs4);
vs5 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep*5), v, vs5);
vs6 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep*6), v, vs6);
vs7 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr + wstep*7), v, vs7);
}
if (k != vecsize) {
// Tail
k = vecsize - 8;
wptr = weights + i * wstep + k;
__m256 v = _mm256_loadu_ps(vec + k);
v = _mm256_and_ps(v, tailMask);
vs0 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr), tailMask), v, vs0);
vs1 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep), tailMask), v, vs1);
vs2 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep * 2), tailMask), v, vs2);
vs3 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep * 3), tailMask), v, vs3);
vs4 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep * 4), tailMask), v, vs4);
vs5 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep * 5), tailMask), v, vs5);
vs6 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep * 6), tailMask), v, vs6);
vs7 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr + wstep * 7), tailMask), v, vs7);
}
__m256 s0 = _mm256_hadd_ps(_mm256_hadd_ps(vs0, vs1), _mm256_hadd_ps(vs2, vs3));
@ -598,10 +627,20 @@ void fastGEMM1T( const float* vec, const float* weights,
const float* wptr = weights + i*wstep;
__m256 vs0 = _mm256_setzero_ps();
for( int k = 0; k < vecsize; k += 8, wptr += 8 )
int k = 0;
for( ; k <= vecsize-8; k += 8, wptr += 8 )
{
__m256 v = _mm256_load_ps(vec + k);
vs0 = _mm256_fmadd_ps(_mm256_load_ps(wptr), v, vs0);
__m256 v = _mm256_loadu_ps(vec + k);
vs0 = _mm256_fmadd_ps(_mm256_loadu_ps(wptr), v, vs0);
}
if (k != vecsize) {
// Tail
k = vecsize - 8;
wptr = weights + i * wstep + k;
__m256 v = _mm256_loadu_ps(vec + k);
v = _mm256_and_ps(v, tailMask);
vs0 = _mm256_fmadd_ps(_mm256_and_ps(_mm256_loadu_ps(wptr), tailMask), v, vs0);
}
__m256 s0 = _mm256_hadd_ps(_mm256_hadd_ps(vs0, vs0), vs0);

2
modules/dnn/src/layers/lrn_layer.cpp
Unescape View File

@ -87,7 +87,7 @@ public:
if (size % 2 != 1 || size <= 0)
CV_Error(Error::StsBadArg, "LRN layer supports only positive odd values for local_size");
alpha = params.get<double>("alpha", 1);
alpha = params.get<double>("alpha", 0.0001);
beta = params.get<double>("beta", 0.75);
bias = params.get<double>("bias", 1);
normBySize = params.get<bool>("norm_by_size", true);

103
modules/dnn/src/layers/recurrent_layers.cpp
Unescape View File

@ -46,6 +46,8 @@
#include <cmath>
#include <opencv2/dnn/shape_utils.hpp>
#include "layers_common.hpp"
namespace cv
{
namespace dnn
@ -118,10 +120,23 @@ class LSTMLayerImpl CV_FINAL : public LSTMLayer
ActivationFunction g_activation;
ActivationFunction h_activation;
#if CV_TRY_AVX
bool useAVX;
#endif
#if CV_TRY_AVX2
bool useAVX2;
#endif
public:
LSTMLayerImpl(const LayerParams& params)
: numTimeStamps(0), numSamples(0)
#if CV_TRY_AVX
, useAVX(checkHardwareSupport(CPU_AVX))
#endif
#if CV_TRY_AVX2
, useAVX2(checkHardwareSupport(CPU_AVX2))
#endif
{
setParamsFrom(params);
@ -343,6 +358,15 @@ public:
hOutTs = hOutTs.colRange(i * hOutTs.cols / numDirs, (i + 1) * hOutTs.cols / numDirs);
Mat cOutTs = produceCellOutput ? output[1].reshape(1, numSamplesTotal) : Mat();
#if CV_TRY_AVX2 || CV_TRY_AVX
bool canUseAvx = gates.isContinuous() && bias.isContinuous()
&& Wx.depth() == CV_32F && gates.depth() == CV_32F
&& bias.depth() == CV_32F && Wx.cols >= 8;
bool canUseAvx_hInternal = hInternal.isContinuous() && gates.isContinuous() && bias.isContinuous()
&& Wh.depth() == CV_32F && hInternal.depth() == CV_32F && gates.depth() == CV_32F
&& Wh.cols >= 8;
#endif
int tsStart, tsEnd, tsInc;
if (reverse || i == 1) {
tsStart = numTimeStamps - 1;
@ -359,9 +383,82 @@ public:
Range curRowRange(ts*numSamples, (ts + 1)*numSamples);
Mat xCurr = xTs.rowRange(curRowRange);
gemm(xCurr, Wx, 1, gates, 0, gates, GEMM_2_T); // Wx * x_t
gemm(hInternal, Wh, 1, gates, 1, gates, GEMM_2_T); //+Wh * h_{t-1}
gemm(dummyOnes, bias, 1, gates, 1, gates); //+b
#if CV_TRY_AVX2
if (useAVX2 && canUseAvx && xCurr.isContinuous())
{
for (int n = 0; n < xCurr.rows; n++) {
opt_AVX2::fastGEMM1T(
xCurr.ptr<float>(n),
Wx.ptr<float>(),
Wx.step1(),
bias.ptr<float>(),
gates.ptr<float>(n),
Wx.rows,
Wx.cols
);
}
}
else
#endif
#if CV_TRY_AVX
if (useAVX && canUseAvx && xCurr.isContinuous())
{
for (int n = 0; n < xCurr.rows; n++) {
opt_AVX::fastGEMM1T(
xCurr.ptr<float>(n),
Wx.ptr<float>(),
Wx.step1(),
bias.ptr<float>(),
gates.ptr<float>(n),
Wx.rows,
Wx.cols
);
}
}
else
#endif
{
gemm(xCurr, Wx, 1, gates, 0, gates, GEMM_2_T); // Wx * x_t
gemm(dummyOnes, bias, 1, gates, 1, gates); //+b
}
#if CV_TRY_AVX2
if (useAVX2 && canUseAvx_hInternal)
{
for (int n = 0; n < hInternal.rows; n++) {
opt_AVX2::fastGEMM1T(
hInternal.ptr<float>(n),
Wh.ptr<float>(),
Wh.step1(),
gates.ptr<float>(n),
gates.ptr<float>(n),
Wh.rows,
Wh.cols
);
}
}
else
#endif
#if CV_TRY_AVX
if (useAVX && canUseAvx_hInternal)
{
for (int n = 0; n < hInternal.rows; n++) {
opt_AVX::fastGEMM1T(
hInternal.ptr<float>(n),
Wh.ptr<float>(),
Wh.step1(),
gates.ptr<float>(n),
gates.ptr<float>(n),
Wh.rows,
Wh.cols
);
}
}
else
#endif
{
gemm(hInternal, Wh, 1, gates, 1, gates, GEMM_2_T); //+Wh * h_{t-1}
}
Mat gateI = gates.colRange(0*numOut, 1*numOut);
Mat gateF = gates.colRange(1*numOut, 2*numOut);

1
modules/dnn/src/ocl4dnn/src/ocl4dnn_conv_spatial.cpp
Unescape View File

@ -48,7 +48,6 @@
#include <vector>
#include <fstream>
#include <sys/stat.h>
#include <assert.h>
#include "../include/common.hpp"
#include "../include/ocl4dnn.hpp"
#include "opencl_kernels_dnn.hpp"

70
modules/dnn/src/onnx/onnx_graph_simplifier.cpp
Unescape View File

@ -110,17 +110,10 @@ private:
opencv_onnx::GraphProto& net;
};
class SoftMaxSubgraph : public Subgraph
class SoftMaxSubgraphBase : public Subgraph
{
public:
SoftMaxSubgraph() : axis(1)
{
int input = addNodeToMatch("");
int inpExp = addNodeToMatch("Exp", input);
int sum = addNodeToMatch("ReduceSum", inpExp);
addNodeToMatch("Div", inpExp, sum);
setFusedNode("Softmax", input);
}
SoftMaxSubgraphBase() : axis(1), id(-1) {}
virtual bool match(const Ptr<ImportGraphWrapper>& net, int nodeId,
std::vector<int>& matchedNodesIds,
@ -128,7 +121,8 @@ public:
{
if (Subgraph::match(net, nodeId, matchedNodesIds, targetNodesIds))
{
Ptr<ImportNodeWrapper> sum = net->getNode(matchedNodesIds[1]);
CV_Assert(id >= 0 && id < matchedNodesIds.size());
Ptr<ImportNodeWrapper> sum = net->getNode(matchedNodesIds[id]);
opencv_onnx::NodeProto* node = sum.dynamicCast<ONNXNodeWrapper>()->node;
for (int i = 0; i < node->attribute_size(); i++)
@ -156,8 +150,60 @@ public:
attr->set_i(axis);
}
private:
protected:
int axis;
int id;
};
class SoftMaxSubgraph : public SoftMaxSubgraphBase
{
public:
SoftMaxSubgraph()
{
int input = addNodeToMatch("");
int inpExp = addNodeToMatch("Exp", input);
int sum = addNodeToMatch("ReduceSum", inpExp);
id = 1;
addNodeToMatch("Div", inpExp, sum);
setFusedNode("Softmax", input);
}
};
class SoftMaxSubgraph2 : public SoftMaxSubgraphBase {
public:
SoftMaxSubgraph2() {
int input = addNodeToMatch("");
int reducemax = addNodeToMatch("ReduceMax", input);
id = 0;
int sub = addNodeToMatch("Sub", input, reducemax);
int exp = addNodeToMatch("Exp", sub);
int reducesum = addNodeToMatch("ReduceSum", exp, addNodeToMatch(""));
addNodeToMatch("Div", exp, reducesum);
setFusedNode("Softmax", input);
}
};
class LogSoftMaxSubgraph : public SoftMaxSubgraphBase
{
public:
LogSoftMaxSubgraph()
{
int input = addNodeToMatch("");
int reducemax = addNodeToMatch("ReduceMax", input);
id = 0;
int sub_1 = addNodeToMatch("Sub", input, reducemax);
int exp = addNodeToMatch("Exp", sub_1);
int reducesum = addNodeToMatch("ReduceSum", exp, addNodeToMatch(""));
int log = addNodeToMatch("Log", reducesum);
addNodeToMatch("Sub", sub_1, log);
setFusedNode("LogSoftmax", input);
}
};
class NormalizeSubgraphBase : public Subgraph
@ -577,6 +623,8 @@ void simplifySubgraphs(opencv_onnx::GraphProto& net)
subgraphs.push_back(makePtr<ResizeSubgraph1>());
subgraphs.push_back(makePtr<ResizeSubgraph2>());
subgraphs.push_back(makePtr<SoftMaxSubgraph>());
subgraphs.push_back(makePtr<SoftMaxSubgraph2>());
subgraphs.push_back(makePtr<LogSoftMaxSubgraph>());
subgraphs.push_back(makePtr<NormalizeSubgraph1>());
subgraphs.push_back(makePtr<NormalizeSubgraph2>());
subgraphs.push_back(makePtr<NormalizeSubgraph2_2>());

42
modules/dnn/src/onnx/onnx_importer.cpp
Unescape View File

@ -768,12 +768,25 @@ void ONNXImporter::handleNode(const opencv_onnx::NodeProto& node_proto)
}
}
void setCeilMode(LayerParams& layerParams)
{
// auto_pad attribute is deprecated and uses ceil
if (layerParams.has("pad_mode"))
{
layerParams.set("ceil_mode", true);
}
else if (!layerParams.has("ceil_mode"))
{
layerParams.set("ceil_mode", false);
}
}
void ONNXImporter::parseMaxPool(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto)
{
int depth = layerParams.get<int>("depth", CV_32F);
layerParams.type = (depth == CV_8S) ? "PoolingInt8" : "Pooling";
layerParams.set("pool", "MAX");
layerParams.set("ceil_mode", layerParams.has("pad_mode"));
setCeilMode(layerParams);
addLayer(layerParams, node_proto);
}
@ -781,7 +794,7 @@ void ONNXImporter::parseAveragePool(LayerParams& layerParams, const opencv_onnx:
{
layerParams.type = "Pooling";
layerParams.set("pool", "AVE");
layerParams.set("ceil_mode", layerParams.has("pad_mode"));
setCeilMode(layerParams);
layerParams.set("ave_pool_padded_area", framework_name == "pytorch");
addLayer(layerParams, node_proto);
}
@ -1077,6 +1090,7 @@ void ONNXImporter::parseSplit(LayerParams& layerParams, const opencv_onnx::NodeP
}
int depth = layerParams.get<int>("depth", CV_32F);
layerParams.type = (depth == CV_8S) ? "SliceInt8" : "Slice";
layerParams.set("axis", layerParams.get<float>("axis", 0));
addLayer(layerParams, node_proto);
}
@ -1085,6 +1099,14 @@ void ONNXImporter::parseBias(LayerParams& layerParams, const opencv_onnx::NodePr
opencv_onnx::NodeProto node_proto = node_proto_;
const std::string& layer_type = node_proto.op_type();
bool isSub = layer_type == "Sub";
if (layer_type == "Sum" && node_proto.input_size() == 1)
{
layerParams.type = "Identity";
addLayer(layerParams, node_proto);
return;
}
CV_Assert((node_proto.input_size() == 2) || (layer_type == "Sum" && node_proto.input_size() > 2));
if (layer_type == "Sum" && node_proto.input_size() > 2)
@ -1382,15 +1404,15 @@ void ONNXImporter::parseImageScaler(LayerParams& layerParams, const opencv_onnx:
void ONNXImporter::parseClip(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto)
{
layerParams.type = "ReLU6";
replaceLayerParam(layerParams, "min", "min_value");
replaceLayerParam(layerParams, "max", "max_value");
layerParams.set("min_value", layerParams.get<float>("min", -FLT_MAX));
layerParams.set("max_value", layerParams.get<float>("max", FLT_MAX));
addLayer(layerParams, node_proto);
}
void ONNXImporter::parseLeakyRelu(LayerParams& layerParams, const opencv_onnx::NodeProto& node_proto)
{
layerParams.type = "ReLU";
replaceLayerParam(layerParams, "alpha", "negative_slope");
layerParams.set("negative_slope", layerParams.get<float>("alpha", 0.01));
addLayer(layerParams, node_proto);
}
@ -1875,6 +1897,16 @@ void ONNXImporter::parseTranspose(LayerParams& layerParams, const opencv_onnx::N
int depth = layerParams.get<int>("depth", CV_32F);
layerParams.type = (depth == CV_8S) ? "PermuteInt8" : "Permute";
replaceLayerParam(layerParams, "perm", "order");
if (!layerParams.has("order")) {
MatShape inpShape = outShapes[node_proto.input(0)];
size_t dims = inpShape.size();
std::vector<int> perm(dims);
for (size_t d = 0; d < dims; ++d)
{
perm[d] = static_cast<int>(dims - 1 - d);
}
layerParams.set("order", DictValue::arrayInt(perm.data(), perm.size()));
}
CV_Assert(node_proto.input_size() == 1);
if (constBlobs.find(node_proto.input(0)) != constBlobs.end())

5
modules/dnn/src/opencl/activations.cl
Unescape View File

@ -131,13 +131,14 @@ __kernel void PowForward(const int n, __global const T* in, __global T* out,
out[index] = pow(shift + scale * in[index], power);
}
__kernel void ELUForward(const int n, __global const T* in, __global T* out)
__kernel void ELUForward(const int n, __global const T* in, __global T* out,
const KERNEL_ARG_DTYPE alpha)
{
int index = get_global_id(0);
if (index < n)
{
T src = in[index];
out[index] = (src >= 0.f) ? src : exp(src) - 1;
out[index] = (src >= 0.f) ? src : alpha * (exp(src) - 1);
}
}

2
modules/dnn/test/test_backends.cpp
Unescape View File

@ -274,7 +274,7 @@ TEST_P(DNNTestNetwork, MobileNet_SSD_v1_TensorFlow_Different_Width_Height)
{
if (backend == DNN_BACKEND_HALIDE)
applyTestTag(CV_TEST_TAG_DNN_SKIP_HALIDE);
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 || backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) &&
target == DNN_TARGET_MYRIAD && getInferenceEngineVPUType() == CV_DNN_INFERENCE_ENGINE_VPU_TYPE_MYRIAD_X)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD_X);

24
modules/dnn/test/test_caffe_importer.cpp
Unescape View File

@ -112,10 +112,12 @@ TEST(Test_Caffe, read_googlenet)
TEST_P(Test_Caffe_nets, Axpy)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
String proto = _tf("axpy.prototxt");
Net net = readNetFromCaffe(proto);
@ -156,7 +158,12 @@ TEST_P(Test_Caffe_nets, Axpy)
l1 = 2e-4;
lInf = 1e-3;
}
else if(target == DNN_TARGET_CUDA_FP16)
if (target == DNN_TARGET_MYRIAD)
{
l1 = 0.001;
lInf = 0.001;
}
if(target == DNN_TARGET_CUDA_FP16)
{
l1 = 0.0002;
lInf = 0.0007;
@ -688,7 +695,7 @@ TEST_P(Test_Caffe_nets, FasterRCNN_vgg16)
CV_TEST_TAG_DEBUG_VERYLONG
);
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 || backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);
@ -699,6 +706,19 @@ TEST_P(Test_Caffe_nets, FasterRCNN_vgg16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE exception: Ngraph operation Reshape with name rpn_cls_score_reshape has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
// Check 'backward_compatible_check || in_out_elements_equal' failed at core/src/op/reshape.cpp:390:
// While validating node 'v1::Reshape bbox_pred_reshape (bbox_pred[0]:f32{1,84}, Constant_241202[0]:i64{4}) -> (f32{?,?,?,?})' with friendly_name 'bbox_pred_reshape':
// Requested output shape {1,6300,4,1} is incompatible with input shape Shape{1, 84}
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
static Mat ref = (Mat_<float>(3, 7) << 0, 2, 0.949398, 99.2454, 210.141, 601.205, 462.849,
0, 7, 0.997022, 481.841, 92.3218, 722.685, 175.953,
0, 12, 0.993028, 133.221, 189.377, 350.994, 563.166);

70
modules/dnn/test/test_darknet_importer.cpp
Unescape View File

@ -121,7 +121,7 @@ public:
{
SCOPED_TRACE("batch size 2");
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (target == DNN_TARGET_MYRIAD && name == "shortcut")
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD);
#endif
@ -442,21 +442,30 @@ TEST_P(Test_Darknet_nets_async, Accuracy)
{
Backend backendId = get<0>(get<1>(GetParam()));
Target targetId = get<1>(get<1>(GetParam()));
std::string prefix = get<0>(GetParam());
applyTestTag(CV_TEST_TAG_MEMORY_512MB);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (INF_ENGINE_VER_MAJOR_LT(2019020000) && backendId == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
applyTestTag(CV_TEST_TAG_MEMORY_512MB);
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
std::string prefix = get<0>(GetParam());
if (backendId != DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && backendId != DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
throw SkipTestException("No support for async forward");
#if defined(INF_ENGINE_RELEASE)
#if INF_ENGINE_VER_MAJOR_GE(2021040000)
if (targetId == DNN_TARGET_MYRIAD && prefix == "yolov3") // NC_OUT_OF_MEMORY
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#else
if (targetId == DNN_TARGET_MYRIAD && prefix == "yolov4") // NC_OUT_OF_MEMORY
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
if (backendId != DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && backendId != DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
throw SkipTestException("No support for async forward");
#endif
#endif
const int numInputs = 2;
std::vector<Mat> inputs(numInputs);
@ -485,6 +494,34 @@ TEST_P(Test_Darknet_nets_async, Accuracy)
netAsync.setPreferableBackend(backendId);
netAsync.setPreferableTarget(targetId);
double l1 = 0.0;
double lInf = 0.0;
#if defined(INF_ENGINE_RELEASE)
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
if (targetId == DNN_TARGET_MYRIAD && prefix == "yolo-voc")
{
l1 = 0.02;
lInf = 0.15;
}
if (targetId == DNN_TARGET_OPENCL_FP16 && prefix == "yolo-voc")
{
l1 = 0.02;
lInf = 0.1;
}
if (targetId == DNN_TARGET_OPENCL_FP16 && prefix == "yolov3")
{
l1 = 0.001;
lInf = 0.007;
}
if (targetId == DNN_TARGET_OPENCL_FP16 && prefix == "yolov4")
{
l1 = 0.001;
lInf = 0.005;
}
}
#endif
// Run asynchronously. To make test more robust, process inputs in the reversed order.
for (int i = numInputs - 1; i >= 0; --i)
{
@ -494,7 +531,7 @@ TEST_P(Test_Darknet_nets_async, Accuracy)
ASSERT_TRUE(out.valid());
Mat result;
EXPECT_TRUE(out.get(result, async_timeout));
normAssert(refs[i], result, format("Index: %d", i).c_str(), 0, 0);
normAssert(refs[i], result, format("Index: %d", i).c_str(), l1, lInf);
}
}
@ -870,10 +907,23 @@ TEST_P(Test_Darknet_layers, avgpool_softmax)
TEST_P(Test_Darknet_layers, region)
{
#if defined(INF_ENGINE_RELEASE)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && INF_ENGINE_VER_MAJOR_GE(2020020000))
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && INF_ENGINE_VER_MAJOR_GE(2020020000))
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy on CPU, OpenCL
// Expected: (normInf) <= (lInf), actual: 0.763223 vs 0.0001
// |ref| = 1.207319974899292
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_CPU, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
testDarknetLayer("region");
}

25
modules/dnn/test/test_halide_layers.cpp
Unescape View File

@ -243,9 +243,11 @@ TEST_P(LRN, Accuracy)
Backend backendId = get<0>(get<5>(GetParam()));
Target targetId = get<1>(get<5>(GetParam()));
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if ((inSize.width == 5 || inSize.height == 5) && targetId == DNN_TARGET_MYRIAD &&
nrmType == "ACROSS_CHANNELS")
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD);
#endif
LayerParams lp;
lp.set("norm_region", nrmType);
@ -410,12 +412,14 @@ TEST_P(FullyConnected, Accuracy)
bool hasBias = get<3>(GetParam());
Backend backendId = get<0>(get<4>(GetParam()));
Target targetId = get<1>(get<4>(GetParam()));
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if ((backendId == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 ||
backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) && (targetId == DNN_TARGET_OPENCL_FP16 ||
(targetId == DNN_TARGET_MYRIAD && getInferenceEngineVPUType() == CV_DNN_INFERENCE_ENGINE_VPU_TYPE_MYRIAD_X))) {
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD_X);
}
#endif
Mat weights(outChannels, inChannels * inSize.height * inSize.width, CV_32F);
randu(weights, -1.0f, 1.0f);
@ -435,9 +439,22 @@ TEST_P(FullyConnected, Accuracy)
Mat input(4, &sz[0], CV_32F);
double l1 = 0.0;
double lInf = 0.0;
#if defined(INF_ENGINE_RELEASE)
if (targetId == DNN_TARGET_MYRIAD)
{
l1 = 0.015;
lInf = 0.025;
}
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && targetId == DNN_TARGET_OPENCL_FP16)
{
l1 = 0.01;
}
#endif
if (targetId == DNN_TARGET_CUDA_FP16)
l1 = 0.015;
test(lp, input, backendId, targetId, false, true, l1);
test(lp, input, backendId, targetId, false, l1, lInf);
}
INSTANTIATE_TEST_CASE_P(Layer_Test_Halide, FullyConnected, Combine(
@ -819,18 +836,18 @@ TEST_P(Eltwise, Accuracy)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019010000)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019010000) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && numConv > 1)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && targetId == DNN_TARGET_OPENCL &&
op == "sum" && numConv == 1 && !weighted)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
#endif
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && numConv > 1)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif

68
modules/dnn/test/test_layers.cpp
Unescape View File

@ -196,13 +196,23 @@ TEST_P(Test_Caffe_layers, DeConvolution)
TEST_P(Test_Caffe_layers, InnerProduct)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// IE exception: Ngraph operation Reshape with name Reshape_4219609 has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
#endif
if (backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
testLayerUsingCaffeModels("layer_inner_product", true);
}
@ -300,10 +310,12 @@ TEST_P(Test_Caffe_layers, Concat)
CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH &&
(target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16))
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#endif
testLayerUsingCaffeModels("layer_concat");
@ -1434,62 +1446,6 @@ INSTANTIATE_TEST_CASE_P(/*nothing*/, Test_DLDT_two_inputs_3dim, Combine(
testing::ValuesIn(list_sizes)
));
typedef testing::TestWithParam<tuple<int, int, tuple<Backend, Target> > > Test_DLDT_two_inputs;
TEST_P(Test_DLDT_two_inputs, as_backend)
{
static const float kScale = 0.5f;
static const float kScaleInv = 1.0f / kScale;
Backend backendId = get<0>(get<2>(GetParam()));
Target targetId = get<1>(get<2>(GetParam()));
Net net;
LayerParams lp;
lp.type = "Eltwise";
lp.name = "testLayer";
lp.set("operation", "sum");
int eltwiseId = net.addLayerToPrev(lp.name, lp.type, lp); // connect to a first input
net.connect(0, 1, eltwiseId, 1); // connect to a second input
int inpSize[] = {1, 2, 3, 4};
Mat firstInp(4, &inpSize[0], get<0>(GetParam()));
Mat secondInp(4, &inpSize[0], get<1>(GetParam()));
randu(firstInp, 0, 255);
randu(secondInp, 0, 255);
net.setInputsNames({"data", "second_input"});
net.setInput(firstInp, "data", kScale);
net.setInput(secondInp, "second_input", kScaleInv);
net.setPreferableBackend(backendId);
net.setPreferableTarget(targetId);
Mat out = net.forward();
Mat ref;
addWeighted(firstInp, kScale, secondInp, kScaleInv, 0, ref, CV_32F);
// Output values are in range [0, 637.5].
double l1 = (targetId == DNN_TARGET_OPENCL_FP16 || targetId == DNN_TARGET_MYRIAD) ? 0.06 : 1e-6;
double lInf = (targetId == DNN_TARGET_OPENCL_FP16 || targetId == DNN_TARGET_MYRIAD) ? 0.3 : 1e-5;
if (targetId == DNN_TARGET_CUDA_FP16)
{
l1 = 0.06;
lInf = 0.3;
}
normAssert(out, ref, "", l1, lInf);
if (cvtest::debugLevel > 0 || HasFailure())
{
std::cout << "input1 scale=" << kScale << " input2 scale=" << kScaleInv << std::endl;
std::cout << "input1: " << firstInp.size << " " << firstInp.reshape(1, 1) << std::endl;
std::cout << "input2: " << secondInp.size << " " << secondInp.reshape(1, 1) << std::endl;
std::cout << "ref: " << ref.reshape(1, 1) << std::endl;
std::cout << "out: " << out.reshape(1, 1) << std::endl;
}
}
INSTANTIATE_TEST_CASE_P(/*nothing*/, Test_DLDT_two_inputs, Combine(
Values(CV_8U, CV_32F), Values(CV_8U, CV_32F),
dnnBackendsAndTargets()
));
class UnsupportedLayer : public Layer
{
public:

63
modules/dnn/test/test_misc.cpp
Unescape View File

@ -582,7 +582,8 @@ TEST_P(Async, create_layer_pipeline_set_and_forward_all)
if (backendId != DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && backendId != DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
throw SkipTestException("No support for async forward");
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
// Exception: Default implementation fallbacks in asynchronous mode
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && dtype == CV_8U)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
if (backendId == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
@ -832,4 +833,64 @@ INSTANTIATE_TEST_CASE_P(/**/, Test_Model_Optimizer,
#endif // HAVE_INF_ENGINE
typedef testing::TestWithParam<tuple<MatDepth, MatDepth, tuple<Backend, Target> > > Test_two_inputs;
TEST_P(Test_two_inputs, basic)
{
static const float kScale = 0.5f;
static const float kScaleInv = 1.0f / kScale;
Backend backendId = get<0>(get<2>(GetParam()));
Target targetId = get<1>(get<2>(GetParam()));
Net net;
LayerParams lp;
lp.type = "Eltwise";
lp.name = "testLayer";
lp.set("operation", "sum");
int eltwiseId = net.addLayerToPrev(lp.name, lp.type, lp); // connect to a first input
net.connect(0, 1, eltwiseId, 1); // connect to a second input
int inpSize[] = {1, 2, 3, 4};
Mat firstInp(4, &inpSize[0], get<0>(GetParam()));
Mat secondInp(4, &inpSize[0], get<1>(GetParam()));
randu(firstInp, 0, 100);
randu(secondInp, 0, 100);
#ifndef CV_CXX11
std::vector<String> input_names;
input_names.push_back("data");
input_names.push_back("second_input");
net.setInputsNames(input_names);
#else
net.setInputsNames({"data", "second_input"});
#endif
net.setInput(firstInp, "data", kScale);
net.setInput(secondInp, "second_input", kScaleInv);
net.setPreferableBackend(backendId);
net.setPreferableTarget(targetId);
Mat out = net.forward();
Mat ref;
addWeighted(firstInp, kScale, secondInp, kScaleInv, 0, ref, CV_32F);
double l1 = (targetId == DNN_TARGET_OPENCL_FP16 || targetId == DNN_TARGET_MYRIAD) ? 0.06 : 1e-6;
double lInf = (targetId == DNN_TARGET_OPENCL_FP16 || targetId == DNN_TARGET_MYRIAD) ? 0.3 : 1e-5;
normAssert(out, ref, "", l1, lInf);
if (cvtest::debugLevel > 0 || HasFailure())
{
std::cout << "input1 scale=" << kScale << " input2 scale=" << kScaleInv << std::endl;
std::cout << "input1: " << firstInp.size << " " << firstInp.reshape(1, 1) << std::endl;
std::cout << "input2: " << secondInp.size << " " << secondInp.reshape(1, 1) << std::endl;
std::cout << "ref: " << ref.reshape(1, 1) << std::endl;
std::cout << "out: " << out.reshape(1, 1) << std::endl;
}
}
INSTANTIATE_TEST_CASE_P(/*nothing*/, Test_two_inputs, Combine(
Values(CV_32F, CV_8U),
Values(CV_32F, CV_8U),
dnnBackendsAndTargets()
));
}} // namespace

210
modules/dnn/test/test_onnx_importer.cpp
Unescape View File

@ -197,17 +197,11 @@ TEST_P(Test_ONNX_layers, Gather)
TEST_P(Test_ONNX_layers, Convolution3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
testONNXModels("conv3d");
}
TEST_P(Test_ONNX_layers, Convolution3D_bias)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
testONNXModels("conv3d_bias");
}
@ -235,18 +229,73 @@ TEST_P(Test_ONNX_layers, Deconvolution)
TEST_P(Test_ONNX_layers, Deconvolution3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2018050000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// ok
// [ GENERAL_ERROR ] vpu/graph_transformer/src/frontend/frontend.cpp:439 Failed to compile layer "2":
// [ GENERAL_ERROR ] vpu/graph_transformer/src/model/model.cpp:198 duplicateData error: while duplicating 2@weights Const data got different desc and content byte sizes (162 and 486 respectively)
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
else if (backend == DNN_BACKEND_OPENCV || target != DNN_TARGET_CPU)
throw SkipTestException("Only DLIE backend on CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV)
throw SkipTestException("OpenCV backend is not supported"); // FIXIT use tags
testONNXModels("deconv3d");
}
TEST_P(Test_ONNX_layers, Deconvolution3D_bias)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// [ GENERAL_ERROR ] vpu/graph_transformer/src/frontend/frontend.cpp:439 Failed to compile layer "2":
// [ GENERAL_ERROR ] vpu/graph_transformer/src/model/model.cpp:198 duplicateData error: while duplicating 2@weights Const data got different desc and content byte sizes (162 and 486 respectively)
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
#endif
if (backend == DNN_BACKEND_OPENCV)
throw SkipTestException("OpenCV backend is not supported"); // FIXIT use tags
testONNXModels("deconv3d_bias");
}
TEST_P(Test_ONNX_layers, Deconvolution3D_pad)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// [ GENERAL_ERROR ] vpu/graph_transformer/src/frontend/frontend.cpp:439 Failed to compile layer "2":
// [ GENERAL_ERROR ] vpu/graph_transformer/src/model/model.cpp:198 duplicateData error: while duplicating 2@weights Const data got different desc and content byte sizes (162 and 486 respectively)
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
#endif
if (backend == DNN_BACKEND_OPENCV)
throw SkipTestException("OpenCV backend is not supported"); // FIXIT use tags
testONNXModels("deconv3d_pad");
}
TEST_P(Test_ONNX_layers, Deconvolution3D_adjpad)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// [ GENERAL_ERROR ] vpu/graph_transformer/src/frontend/frontend.cpp:439 Failed to compile layer "2":
// [ GENERAL_ERROR ] vpu/graph_transformer/src/model/model.cpp:198 duplicateData error: while duplicating 2@weights Const data got different desc and content byte sizes (162 and 486 respectively)
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
#endif
if (backend == DNN_BACKEND_OPENCV)
throw SkipTestException("OpenCV backend is not supported"); // FIXIT use tags
testONNXModels("deconv3d_adjpad");
}
@ -317,16 +366,14 @@ TEST_P(Test_ONNX_layers, Scale)
TEST_P(Test_ONNX_layers, ReduceMean3D)
{
if (backend == DNN_BACKEND_CUDA)
{
// ok
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
testONNXModels("reduce_mean3d");
}
@ -443,13 +490,12 @@ TEST_P(Test_ONNX_layers, Concatenation)
TEST_P(Test_ONNX_layers, Eltwise3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
#endif
testONNXModels("eltwise3d");
}
@ -460,55 +506,56 @@ TEST_P(Test_ONNX_layers, AveragePooling)
TEST_P(Test_ONNX_layers, MaxPooling3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// ok
// accuracy
if (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16)
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
// IE exception: [ GENERAL_ERROR ] AssertionFailed: !expired()
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
testONNXModels("max_pool3d", npy, 0, 0, false, false);
}
TEST_P(Test_ONNX_layers, AvePooling3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
{
// ok
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
testONNXModels("ave_pool3d");
}
TEST_P(Test_ONNX_layers, PoolConv3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
{
// ok
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
testONNXModels("pool_conv_3d");
}
@ -1011,6 +1058,7 @@ TEST_P(Test_ONNX_layers, DynamicAxes)
TEST_P(Test_ONNX_layers, MaxPool1d)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
@ -1019,11 +1067,20 @@ TEST_P(Test_ONNX_layers, MaxPool1d)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
{
// 2021.4: [ GENERAL_ERROR ] AssertionFailed: !expired()
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
testONNXModels("maxpooling_1d");
}
TEST_P(Test_ONNX_layers, MaxPoolSigmoid1d)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
@ -1032,11 +1089,13 @@ TEST_P(Test_ONNX_layers, MaxPoolSigmoid1d)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
testONNXModels("maxpooling_sigmoid_1d");
}
TEST_P(Test_ONNX_layers, MaxPool1d_Twise)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
@ -1045,11 +1104,13 @@ TEST_P(Test_ONNX_layers, MaxPool1d_Twise)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
testONNXModels("two_maxpooling_1d");
}
TEST_P(Test_ONNX_layers, AvePool1d)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
@ -1058,11 +1119,13 @@ TEST_P(Test_ONNX_layers, AvePool1d)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
testONNXModels("average_pooling_1d");
}
TEST_P(Test_ONNX_layers, PoolConv1d)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
@ -1071,6 +1134,7 @@ TEST_P(Test_ONNX_layers, PoolConv1d)
{
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
}
#endif
testONNXModels("pool_conv_1d");
}
@ -1257,11 +1321,18 @@ TEST_P(Test_ONNX_nets, Squeezenet)
TEST_P(Test_ONNX_nets, Googlenet)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
// accuracy
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
const String model = _tf("models/googlenet.onnx", false);
@ -1523,7 +1594,7 @@ TEST_P(Test_ONNX_nets, DenseNet121)
TEST_P(Test_ONNX_nets, Inception_v1)
{
#if defined(INF_ENGINE_RELEASE)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 ||
backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH) && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD);
@ -1533,30 +1604,35 @@ TEST_P(Test_ONNX_nets, Inception_v1)
TEST_P(Test_ONNX_nets, Shufflenet)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
{
if (target == DNN_TARGET_OPENCL_FP16) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (target == DNN_TARGET_OPENCL) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (target == DNN_TARGET_MYRIAD) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
}
#endif
testONNXModels("shufflenet", pb);
}
TEST_P(Test_ONNX_nets, Resnet34_kinetics)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
{
// ok
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// IE exception: Function contains several inputs and outputs with one friendly name!
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
String onnxmodel = findDataFile("dnn/resnet-34_kinetics.onnx", false);
Mat image0 = imread(findDataFile("dnn/dog416.png"));
@ -1595,7 +1671,13 @@ TEST_P(Test_ONNX_nets, Resnet34_kinetics)
net.setPreferableTarget(target);
// output range [-5, 11]
float l1 = 0.0013, lInf = 0.009;
float l1 = 0.0013;
float lInf = 0.009;
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_OPENCL_FP16)
{
l1 = 0.02;
lInf = 0.07;
}
if (target == DNN_TARGET_CUDA_FP16)
{
l1 = 0.01;

103
modules/dnn/test/test_tf_importer.cpp
Unescape View File

@ -216,13 +216,12 @@ TEST_P(Test_TensorFlow_layers, conv_pool_nchw)
TEST_P(Test_TensorFlow_layers, Convolution3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
#endif
runTensorFlowNet("conv3d");
}
@ -231,7 +230,7 @@ TEST_P(Test_TensorFlow_layers, padding)
runTensorFlowNet("padding_valid");
runTensorFlowNet("spatial_padding");
runTensorFlowNet("mirror_pad");
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019020000)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019020000) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (target == DNN_TARGET_MYRIAD)
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
@ -344,6 +343,7 @@ TEST_P(Test_TensorFlow_layers, concat_axis_1)
TEST_P(Test_TensorFlow_layers, concat_3d)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
{
if (target == DNN_TARGET_OPENCL_FP16) applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16);
@ -353,6 +353,7 @@ TEST_P(Test_TensorFlow_layers, concat_3d)
if ((backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH ||
backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019) && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
#endif
runTensorFlowNet("concat_3d");
}
@ -430,22 +431,32 @@ TEST_P(Test_TensorFlow_layers, batch_norm3D)
TEST_P(Test_TensorFlow_layers, slim_batch_norm)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
// Output values range: [-40.0597, 207.827]
double l1 = default_l1, lInf = default_lInf;
double l1 = default_l1;
double lInf = default_lInf;
if (target == DNN_TARGET_OPENCL_FP16 || target == DNN_TARGET_MYRIAD)
{
l1 = 0.041;
lInf = 0.33;
}
#if defined(INF_ENGINE_RELEASE)
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_CPU)
{
lInf = 0.0002;
}
#endif
else if (target == DNN_TARGET_CUDA_FP16)
{
l1 = 0.005;
lInf = 0.33;
}
runTensorFlowNet("slim_batch_norm", false, l1, lInf);
}
@ -572,7 +583,7 @@ TEST_P(Test_TensorFlow_layers, max_pool_grad)
TEST_P(Test_TensorFlow_layers, ave_pool_same)
{
// Reference output values are in range [-0.519531, 0.112976]
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019010000)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_GE(2019010000) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (target == DNN_TARGET_MYRIAD && getInferenceEngineVPUType() == CV_DNN_INFERENCE_ENGINE_VPU_TYPE_MYRIAD_X)
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
@ -586,38 +597,41 @@ TEST_P(Test_TensorFlow_layers, ave_pool_same)
TEST_P(Test_TensorFlow_layers, MaxPooling3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// ok
// accuracy
if (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16)
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
// IE exception: [ GENERAL_ERROR ] AssertionFailed: !expired()
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
runTensorFlowNet("max_pool3d");
}
TEST_P(Test_TensorFlow_layers, AvePooling3D)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2019010000)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
if (backend == DNN_BACKEND_CUDA)
{
// ok
}
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER); // Only CPU on DLIE backend is supported
else if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target != DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH); // Only CPU on DLIE backend is supported
else if (target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported");
#endif
if (backend == DNN_BACKEND_OPENCV && target != DNN_TARGET_CPU)
throw SkipTestException("Only CPU is supported"); // FIXIT use tags
runTensorFlowNet("ave_pool3d");
}
@ -650,10 +664,12 @@ TEST_P(Test_TensorFlow_layers, matmul)
TEST_P(Test_TensorFlow_layers, reshape)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
runTensorFlowNet("shift_reshape_no_reorder");
runTensorFlowNet("reshape_no_reorder");
runTensorFlowNet("reshape_reduce");
@ -1319,18 +1335,35 @@ TEST_P(Test_TensorFlow_layers, lstm)
{
if(backend == DNN_BACKEND_CUDA)
applyTestTag(CV_TEST_TAG_DNN_SKIP_CUDA); /* not supported */
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// Exception: Ngraph operation Reshape with name Reshape has dynamic output shape on 0 port, but CPU plug-in supports only static shape
if (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16)
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
// Xlink
if (target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
#endif
if (backend == DNN_BACKEND_OPENCV && target == DNN_TARGET_OPENCL_FP16)
applyTestTag(CV_TEST_TAG_DNN_SKIP_OPENCL_FP16);
runTensorFlowNet("lstm", true);
runTensorFlowNet("lstm", true, 0.0, 0.0, true);
}
TEST_P(Test_TensorFlow_layers, split)
{
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
@ -1360,8 +1393,10 @@ TEST_P(Test_TensorFlow_layers, resize_nearest_neighbor_align_corners)
TEST_P(Test_TensorFlow_layers, resize_nearest_neighbor_half_pixel)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
runTensorFlowNet("resize_nearest_neighbor", false, 0.0, 0.0, false, "_half_pixel");
}
@ -1500,12 +1535,28 @@ TEST_P(Test_TensorFlow_layers, clip_by_value)
TEST_P(Test_TensorFlow_layers, tf2_prelu)
{
if (backend == DNN_BACKEND_CUDA)
applyTestTag(CV_TEST_TAG_DNN_SKIP_CUDA); // not supported; only across channels is supported
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
if (backend == DNN_BACKEND_CUDA)
applyTestTag(CV_TEST_TAG_DNN_SKIP_CUDA); // not supported; only across channels is supported
#endif
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_EQ(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH)
{
// IE exception: Input prelu:StatefulPartitionedCall/StatefulPartitionedCall/sequential/p_re_lu/add hasn't been found in primitiveIDs map
if (target == DNN_TARGET_OPENCL || target == DNN_TARGET_OPENCL_FP16)
applyTestTag(target == DNN_TARGET_OPENCL ? CV_TEST_TAG_DNN_SKIP_IE_OPENCL : CV_TEST_TAG_DNN_SKIP_IE_OPENCL_FP16,
CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION
);
// IE exception: Eltwise node with name `StatefulPartitionedCall/StatefulPartitionedCall/sequential/p_re_lu/add` has invalid input/output dims configuration
if (target == DNN_TARGET_CPU)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_CPU, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
}
#endif
runTensorFlowNet("tf2_prelu");
}

11
modules/dnn/test/test_torch_importer.cpp
Unescape View File

@ -236,23 +236,32 @@ TEST_P(Test_Torch_layers, net_lp_pooling_square)
}
TEST_P(Test_Torch_layers, net_lp_pooling_power)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH, CV_TEST_TAG_DNN_SKIP_IE_VERSION);
#endif
runTorchNet("net_lp_pooling_power", "", false, true);
}
TEST_P(Test_Torch_layers, net_conv_gemm_lrn)
{
#if defined(INF_ENGINE_RELEASE) && INF_ENGINE_VER_MAJOR_LT(2021040000)
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NN_BUILDER_2019 && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NN_BUILDER);
if (backend == DNN_BACKEND_INFERENCE_ENGINE_NGRAPH && target == DNN_TARGET_MYRIAD)
applyTestTag(CV_TEST_TAG_DNN_SKIP_IE_MYRIAD, CV_TEST_TAG_DNN_SKIP_IE_NGRAPH);
#endif
double l1 = 0.0, lInf = 0.0;
if (target == DNN_TARGET_OPENCL_FP16)
{
l1 = 0.046;
lInf = 0.023;
}
else if (target == DNN_TARGET_MYRIAD)
{
l1 = 0.02;
lInf = 0.05;
}
else if (target == DNN_TARGET_CUDA_FP16)
{
l1 = 0.0042;

2
modules/features2d/test/test_nearestneighbors.cpp
Unescape View File

@ -312,7 +312,7 @@ void CV_FlannSavedIndexTest::createModel(const cv::Mat &data)
case 1: createIndex( data, KDTreeIndexParams() ); break;
//case 2: createIndex( data, CompositeIndexParams() ); break; // nothing to save for linear search
//case 2: createIndex( data, AutotunedIndexParams() ); break; // possible linear index !
default: assert(0);
default: CV_Assert(0);
}
string filename = tempfile();
index->save( filename );

4
modules/flann/include/opencv2/flann/nn_index.h
Unescape View File

@ -106,8 +106,8 @@ public:
fprintf(stderr, "I can only search one feature at a time for range search\n");
return -1;
}
assert(query.cols == veclen());
assert(indices.cols == dists.cols);
CV_Assert(query.cols == veclen());
CV_Assert(indices.cols == dists.cols);
int n = 0;
int* indices_ptr = NULL;

1
modules/highgui/src/window_winrt.cpp
Unescape View File

@ -30,7 +30,6 @@
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include <assert.h>
#include <opencv2\highgui.hpp>
#include <opencv2\highgui\highgui_winrt.hpp>
#include "window_winrt_bridge.hpp"

22
modules/imgproc/src/approx.cpp
Unescape View File

@ -119,7 +119,7 @@ CvSeq* icvApproximateChainTC89( CvChain* chain, int header_size,
}
}
//assert( pt.x == chain->origin.x && pt.y == chain->origin.y );
//CV_Assert( pt.x == chain->origin.x && pt.y == chain->origin.y );
if( method <= CV_CHAIN_APPROX_SIMPLE )
return cvEndWriteSeq( &writer );
@ -129,7 +129,7 @@ CvSeq* icvApproximateChainTC89( CvChain* chain, int header_size,
len = i;
current = temp.next;
assert( current );
CV_Assert( current );
/* Pass 1.
Determines support region for all the remained points */
@ -148,7 +148,7 @@ CvSeq* icvApproximateChainTC89( CvChain* chain, int header_size,
int dx, dy;
Cv32suf d;
assert( k <= len );
CV_Assert( k <= len );
/* calc indices */
i1 = i - k;
@ -205,7 +205,7 @@ CvSeq* icvApproximateChainTC89( CvChain* chain, int header_size,
((double)dx2 * dx2 + (double)dy2 * dy2) ));
sk.f = (float) (temp_num + 1.1);
assert( 0 <= sk.f && sk.f <= 2.2 );
CV_Assert( 0 <= sk.f && sk.f <= 2.2 );
if( j < k && sk.i <= s )
break;
@ -258,7 +258,7 @@ CvSeq* icvApproximateChainTC89( CvChain* chain, int header_size,
/* Pass 3.
Removes non-dominant points with 1-length support region */
current = temp.next;
assert( current );
CV_Assert( current );
prev_current = &temp;
do
@ -293,7 +293,7 @@ CvSeq* icvApproximateChainTC89( CvChain* chain, int header_size,
/* Pass 4.
Cleans remained couples of points */
assert( temp.next );
CV_Assert( temp.next );
if( array[0].s != 0 && array[len - 1].s != 0 ) /* specific case */
{
@ -362,7 +362,7 @@ copy_vect:
// gather points
current = temp.next;
assert( current );
CV_Assert( current );
do
{
@ -439,7 +439,7 @@ cvApproxChains( CvSeq* src_seq,
if( src_seq->v_next && len >= minimal_perimeter )
{
assert( prev_contour != 0 );
CV_Assert( prev_contour != 0 );
parent = prev_contour;
prev_contour = 0;
src_seq = src_seq->v_next;
@ -590,7 +590,7 @@ approxPolyDP_( const Point_<T>* src_contour, int count0, Point_<T>* dst_contour,
dx = end_pt.x - start_pt.x;
dy = end_pt.y - start_pt.y;
assert( dx != 0 || dy != 0 );
CV_Assert( dx != 0 || dy != 0 );
while( pos != slice.end )
{
@ -815,7 +815,7 @@ cvApproxPoly( const void* array, int header_size,
CV_Error( CV_StsBadArg, "Invalid approximation method" );
}
assert( contour );
CV_Assert( contour );
if( header_size >= (int)sizeof(CvContour))
cvBoundingRect( contour, 1 );
@ -836,7 +836,7 @@ cvApproxPoly( const void* array, int header_size,
if( src_seq->v_next )
{
assert( prev_contour != 0 );
CV_Assert( prev_contour != 0 );
parent = prev_contour;
prev_contour = 0;
src_seq = src_seq->v_next;

2
modules/imgproc/src/bilateral_filter.simd.hpp
Unescape View File

@ -205,7 +205,7 @@ public:
}
else
{
assert( cn == 3 );
CV_Assert( cn == 3 );
AutoBuffer<float> buf(alignSize(size.width, CV_SIMD_WIDTH)*3 + size.width + CV_SIMD_WIDTH - 1);
memset(buf.data(), 0, buf.size() * sizeof(float));
float *sum_b = alignPtr(buf.data(), CV_SIMD_WIDTH);

8
modules/imgproc/src/color_lab.cpp
Unescape View File

@ -3567,7 +3567,7 @@ struct Luv2RGBinteger
long long int xv = ((int)up)*(long long)vp;
int x = (int)(xv/BASE);
x = y*x/BASE;
x = ((long long int)y)*x/BASE;
long long int vpl = LUVLUT.LvToVpl_b[LL*256+vv];
long long int zp = vpl - xv*(255/3);
@ -3689,6 +3689,7 @@ struct Luv2RGBinteger
vzm[i] = zm;
vx[i] = (int32_t)(xv >> base_shift);
vx[i] = (((int64_t)y_)*vx[i]) >> base_shift;
}
v_int32 zm[4];
for(int k = 0; k < 4; k++)
@ -3697,11 +3698,6 @@ struct Luv2RGBinteger
zm[k] = vx_load_aligned(vzm + k*vsize/4);
}
for(int k = 0; k < 4; k++)
{
x[k] = (y[k]*x[k]) >> base_shift;
}
// z = zm/256 + zm/65536;
for (int k = 0; k < 4; k++)
{

4
modules/imgproc/src/contours.cpp
Unescape View File

@ -97,7 +97,7 @@ cvReadChainPoint( CvChainPtReader * reader )
reader->ptr = ptr;
reader->code = (schar)code;
assert( (code & ~7) == 0 );
CV_Assert( (code & ~7) == 0 );
reader->pt.x = pt.x + icvCodeDeltas[code].x;
reader->pt.y = pt.y + icvCodeDeltas[code].y;
}
@ -1187,7 +1187,7 @@ cvFindNextContour( CvContourScanner scanner )
}
/* hole flag of the parent must differ from the flag of the contour */
assert( par_info->is_hole != is_hole );
CV_Assert( par_info->is_hole != is_hole );
if( par_info->contour == 0 ) /* removed contour */
goto resume_scan;
}

2
modules/imgproc/src/convhull.cpp
Unescape View File

@ -716,7 +716,7 @@ CV_IMPL CvSeq* cvConvexityDefects( const CvArr* array,
dx0 = (double)hull_next->x - (double)hull_cur->x;
dy0 = (double)hull_next->y - (double)hull_cur->y;
assert( dx0 != 0 || dy0 != 0 );
CV_Assert( dx0 != 0 || dy0 != 0 );
scale = 1./std::sqrt(dx0*dx0 + dy0*dy0);
defect.start = hull_cur;

2
modules/imgproc/src/distransform.cpp
Unescape View File

@ -270,7 +270,7 @@ distanceTransformEx_5x5( const Mat& _src, Mat& _temp, Mat& _dist, Mat& _labels,
if( !s[j] )
{
tmp[j] = 0;
//assert( lls[j] != 0 );
//CV_Assert( lls[j] != 0 );
}
else
{

10
modules/imgproc/src/drawing.cpp
Unescape View File

@ -138,7 +138,7 @@ bool clipLine( Size2l img_size, Point2l& pt1, Point2l& pt2 )
}
}
assert( (c1 & c2) != 0 || (x1 | y1 | x2 | y2) >= 0 );
CV_Assert( (c1 & c2) != 0 || (x1 | y1 | x2 | y2) >= 0 );
}
return (c1 | c2) == 0;
@ -648,7 +648,7 @@ Line2( Mat& img, Point2l pt1, Point2l pt2, const void* color)
size_t step = img.step;
Size size = img.size();
//assert( img && (nch == 1 || nch == 3) && img.depth() == CV_8U );
//CV_Assert( img && (nch == 1 || nch == 3) && img.depth() == CV_8U );
Size2l sizeScaled(((int64)size.width) << XY_SHIFT, ((int64)size.height) << XY_SHIFT);
if( !clipLine( sizeScaled, pt1, pt2 ))
@ -1120,7 +1120,7 @@ FillConvexPoly( Mat& img, const Point2l* v, int npts, const void* color, int lin
p0.x <<= XY_SHIFT - shift;
p0.y <<= XY_SHIFT - shift;
assert( 0 <= shift && shift <= XY_SHIFT );
CV_Assert( 0 <= shift && shift <= XY_SHIFT );
xmin = xmax = v[0].x;
ymin = ymax = v[0].y;
@ -1340,7 +1340,7 @@ FillEdgeCollection( Mat& img, std::vector<PolyEdge>& edges, const void* color )
for( i = 0; i < total; i++ )
{
PolyEdge& e1 = edges[i];
assert( e1.y0 < e1.y1 );
CV_Assert( e1.y0 < e1.y1 );
// Determine x-coordinate of the end of the edge.
// (This is not necessary x-coordinate of any vertex in the array.)
int64 x1 = e1.x + (e1.y1 - e1.y0) * e1.dx;
@ -2613,7 +2613,7 @@ cvDrawContours( void* _img, CvSeq* contour,
char code;
CV_READ_SEQ_ELEM( code, reader );
assert( (code & ~7) == 0 );
CV_Assert( (code & ~7) == 0 );
if( code != prev_code )
{

6
modules/imgproc/src/emd.cpp
Unescape View File

@ -336,7 +336,7 @@ static int icvInitEMD( const float* signature1, int size1,
char *buffer, *buffer_end;
memset( state, 0, sizeof( *state ));
assert( cost_step % sizeof(float) == 0 );
CV_Assert( cost_step % sizeof(float) == 0 );
cost_step /= sizeof(float);
/* calculate buffer size */
@ -510,7 +510,7 @@ static int icvInitEMD( const float* signature1, int size1,
}
else
{
assert( cost );
CV_Assert( cost );
val = cost[cost_step*ci + cj];
}
state->cost[i][j] = val;
@ -552,7 +552,7 @@ static int icvInitEMD( const float* signature1, int size1,
buffer += dsize;
}
assert( buffer <= buffer_end );
CV_Assert( buffer <= buffer_end );
icvRussel( state );

8
modules/imgproc/src/hough.cpp
Unescape View File

@ -340,8 +340,8 @@ HoughLinesSDiv( InputArray image, OutputArray lines, int type,
rv = r0 * std::cos( phi );
i = (int)rv * tn;
i += cvFloor( phi1 );
assert( i >= 0 );
assert( i < rn * tn );
CV_Assert( i >= 0 );
CV_Assert( i < rn * tn );
caccum[i] = (uchar) (caccum[i] + ((i ^ iprev) != 0));
iprev = i;
if( cmax < caccum[i] )
@ -405,8 +405,8 @@ HoughLinesSDiv( InputArray image, OutputArray lines, int type,
i = CV_IMAX( i, -1 );
i = CV_IMIN( i, sfn );
mcaccum[i]++;
assert( i >= -1 );
assert( i <= sfn );
CV_Assert( i >= -1 );
CV_Assert( i <= sfn );
}
}

2
modules/imgproc/src/median_blur.simd.hpp
Unescape View File

@ -463,7 +463,7 @@ medianBlur_8u_Om( const Mat& _src, Mat& _dst, int m )
}
else
{
assert( cn == 4 );
CV_Assert( cn == 4 );
for( k = 0; k < m*4; k += 4 )
{
UPDATE_ACC01( src_top[k], 0, -- );

2
modules/imgproc/src/moments.cpp
Unescape View File

@ -52,7 +52,7 @@ static void completeMomentState( Moments* moments )
double cx = 0, cy = 0;
double mu20, mu11, mu02;
double inv_m00 = 0.0;
assert( moments != 0 );
CV_Assert( moments != 0 );
if( fabs(moments->m00) > DBL_EPSILON )
{

1
modules/imgproc/src/precomp.hpp
Unescape View File

@ -54,7 +54,6 @@
#include "hal_replacement.hpp"
#include <math.h>
#include <assert.h>
#include <string.h>
#include <stdlib.h>
#include <stdio.h>

8
modules/imgproc/src/resize.cpp
Unescape View File

@ -3191,7 +3191,7 @@ static int computeResizeAreaTab( int ssize, int dsize, int cn, double scale, Dec
if( sx1 - fsx1 > 1e-3 )
{
assert( k < ssize*2 );
CV_Assert( k < ssize*2 );
tab[k].di = dx * cn;
tab[k].si = (sx1 - 1) * cn;
tab[k++].alpha = (float)((sx1 - fsx1) / cellWidth);
@ -3199,7 +3199,7 @@ static int computeResizeAreaTab( int ssize, int dsize, int cn, double scale, Dec
for(int sx = sx1; sx < sx2; sx++ )
{
assert( k < ssize*2 );
CV_Assert( k < ssize*2 );
tab[k].di = dx * cn;
tab[k].si = sx * cn;
tab[k++].alpha = float(1.0 / cellWidth);
@ -3207,7 +3207,7 @@ static int computeResizeAreaTab( int ssize, int dsize, int cn, double scale, Dec
if( fsx2 - sx2 > 1e-3 )
{
assert( k < ssize*2 );
CV_Assert( k < ssize*2 );
tab[k].di = dx * cn;
tab[k].si = sx2 * cn;
tab[k++].alpha = (float)(std::min(std::min(fsx2 - sx2, 1.), cellWidth) / cellWidth);
@ -3899,7 +3899,7 @@ void resize(int src_type,
{
if( k == 0 || ytab[k].di != ytab[k-1].di )
{
assert( ytab[k].di == dy );
CV_Assert( ytab[k].di == dy );
tabofs[dy++] = k;
}
}

2
modules/imgproc/src/samplers.cpp
Unescape View File

@ -74,7 +74,7 @@ adjustRect( const uchar* src, size_t src_step, int pix_size,
src += rect.width*pix_size;
rect.width = 0;
}
assert( rect.width <= win_size.width );
CV_Assert( rect.width <= win_size.width );
}
if( ip.y >= 0 )

6
modules/imgproc/src/segmentation.cpp
Unescape View File

@ -155,7 +155,7 @@ void cv::watershed( InputArray _src, InputOutputArray _markers )
dr = std::abs((ptr1)[2] - (ptr2)[2]);\
diff = ws_max(db,dg); \
diff = ws_max(diff,dr); \
assert( 0 <= diff && diff <= 255 ); \
CV_Assert( 0 <= diff && diff <= 255 ); \
}
CV_Assert( src.type() == CV_8UC3 && dst.type() == CV_32SC1 );
@ -215,7 +215,7 @@ void cv::watershed( InputArray _src, InputOutputArray _markers )
}
// Add to according queue
assert( 0 <= idx && idx <= 255 );
CV_Assert( 0 <= idx && idx <= 255 );
ws_push( idx, i*mstep + j, i*istep + j*3 );
m[0] = IN_QUEUE;
}
@ -286,7 +286,7 @@ void cv::watershed( InputArray _src, InputOutputArray _markers )
}
// Set label to current pixel in marker image
assert( lab != 0 );
CV_Assert( lab != 0 );
m[0] = lab;
if( lab == WSHED )

2
modules/imgproc/src/subdivision2d.cpp
Unescape View File

@ -436,7 +436,7 @@ int Subdiv2D::insert(Point2f pt)
else
CV_Error_(CV_StsError, ("Subdiv2D::locate returned invalid location = %d", location) );
assert( curr_edge != 0 );
CV_Assert( curr_edge != 0 );
validGeometry = false;
curr_point = newPoint(pt, false);

4
modules/imgproc/test/test_approxpoly.cpp
Unescape View File

@ -182,7 +182,7 @@ int CV_ApproxPolyTest::check_slice( CvPoint StartPt, CvPoint EndPt,
////////////////////////////////
if( SrcReader == NULL )
{
assert( false );
CV_Assert( false );
return 0;
}
@ -237,7 +237,7 @@ int CV_ApproxPolyTest::check( CvSeq* SrcSeq, CvSeq* DstSeq, float Eps )
int Count;
int i,j;
assert( SrcSeq && DstSeq );
CV_Assert( SrcSeq && DstSeq );
////////// init ////////////////////
Count = SrcSeq->total;

2
modules/imgproc/test/test_bilateral_filter.cpp
Unescape View File

@ -203,7 +203,7 @@ namespace opencv_test { namespace {
}
else
{
assert( cn == 3 );
CV_Assert( cn == 3 );
for( j = 0; j < size.width*3; j += 3 )
{
float sum_b = 0, sum_g = 0, sum_r = 0, wsum = 0;

2
modules/imgproc/test/test_canny.cpp
Unescape View File

@ -249,7 +249,7 @@ test_Canny( const Mat& src, Mat& dst,
}
else
{
assert( fabs(tg) > tan_3pi_8 );
CV_Assert( fabs(tg) > tan_3pi_8 );
x1 = x2 = x; y1 = y + 1; y2 = y - 1;
}

41
modules/imgproc/test/test_color.cpp
Unescape View File

@ -217,7 +217,7 @@ void CV_ColorCvtBaseTest::convert_forward( const Mat& src, Mat& dst )
float* dst_buf = &_dst_buf[0];
int i, j;
assert( (cn == 3 || cn == 4) && (dst_cn == 3 || dst_cn == 1) );
CV_Assert( (cn == 3 || cn == 4) && (dst_cn == 3 || dst_cn == 1) );
for( i = 0; i < src.rows; i++ )
{
@ -281,7 +281,7 @@ void CV_ColorCvtBaseTest::convert_forward( const Mat& src, Mat& dst )
}
break;
default:
assert(0);
CV_Assert(0);
}
}
}
@ -312,7 +312,7 @@ void CV_ColorCvtBaseTest::convert_backward( const Mat& src, const Mat& dst, Mat&
float* dst_buf = &_dst_buf[0];
int i, j;
assert( cn == 3 || cn == 4 );
CV_Assert( cn == 3 || cn == 4 );
for( i = 0; i < src.rows; i++ )
{
@ -385,7 +385,7 @@ void CV_ColorCvtBaseTest::convert_backward( const Mat& src, const Mat& dst, Mat&
}
break;
default:
assert(0);
CV_Assert(0);
}
}
}
@ -1501,7 +1501,7 @@ void CV_ColorRGBTest::convert_forward( const Mat& src, Mat& dst )
int g_rshift = dst_bits == 16 ? 2 : 3;
int r_lshift = dst_bits == 16 ? 11 : 10;
//assert( (cn == 3 || cn == 4) && (dst_cn == 3 || (dst_cn == 2 && depth == CV_8U)) );
//CV_Assert( (cn == 3 || cn == 4) && (dst_cn == 3 || (dst_cn == 2 && depth == CV_8U)) );
for( i = 0; i < src.rows; i++ )
{
@ -1571,7 +1571,7 @@ void CV_ColorRGBTest::convert_forward( const Mat& src, Mat& dst )
}
break;
default:
assert(0);
CV_Assert(0);
}
}
}
@ -1587,7 +1587,7 @@ void CV_ColorRGBTest::convert_backward( const Mat& /*src*/, const Mat& src, Mat&
int g_lshift = dst_bits == 16 ? 2 : 3;
int r_rshift = dst_bits == 16 ? 11 : 10;
//assert( (cn == 3 || cn == 4) && (src_cn == 3 || (src_cn == 2 && depth == CV_8U)) );
//CV_Assert( (cn == 3 || cn == 4) && (src_cn == 3 || (src_cn == 2 && depth == CV_8U)) );
for( i = 0; i < src.rows; i++ )
{
@ -1677,7 +1677,7 @@ void CV_ColorRGBTest::convert_backward( const Mat& /*src*/, const Mat& src, Mat&
}
break;
default:
assert(0);
CV_Assert(0);
}
}
}
@ -1840,6 +1840,21 @@ TEST(Imgproc_ColorLuv, accuracy) { CV_ColorLuvTest test; test.safe_run(); }
TEST(Imgproc_ColorRGB, accuracy) { CV_ColorRGBTest test; test.safe_run(); }
TEST(Imgproc_ColorBayer, accuracy) { CV_ColorBayerTest test; test.safe_run(); }
TEST(Imgproc_ColorLuv, Overflow_21112)
{
const Size sz(107, 16); // unaligned size to run both SIMD and generic code
Mat luv_init(sz, CV_8UC3, Scalar(49, 205, 23));
Mat rgb;
cvtColor(luv_init, rgb, COLOR_Luv2RGB);
// Convert to normal Luv coordinates for floats.
Mat luv_initf(sz, CV_32FC3, Scalar(49.0f/255.f*100, 205.0f*354/255.f - 134, 23.0f*262/255.f - 140));
Mat rgbf;
cvtColor(luv_initf, rgbf, COLOR_Luv2RGB);
Mat rgb_converted;
rgb.convertTo(rgb_converted, CV_32F);
EXPECT_LE(cvtest::norm(255.f*rgbf, rgb_converted, NORM_INF), 1e-5);
}
TEST(Imgproc_ColorBayer, regression)
{
cvtest::TS* ts = cvtest::TS::ptr();
@ -2569,7 +2584,7 @@ int row8uLuv2RGB(const uchar* src_row, uchar *dst_row, int n, int cn, int blue_i
long long int xv = ((int)up)*(long long)vp;
int x = (int)(xv/BASE);
x = y*x/BASE;
x = ((long long int)y)*x/BASE;
long long int vpl = LvToVpl_b[LL*256+vv];
long long int zp = vpl - xv*(255/3);
@ -2725,11 +2740,11 @@ TEST(Imgproc_ColorLuv_Full, bitExactness)
0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0,
0xec311a14, 0x995efefc, 0xf71b590b, 0xc1edfce7, 0x67b2b2e2, 0xe6d7f90d, 0xbcbaff5c, 0xd86ae19c,
0x3e8e4647, 0x53f1a5e3, 0x60dfb6ca, 0xcda851fe, 0xd91084b3, 0xe361bf6f, 0x90fe66ed, 0xb19c5b89,
0x4bff0e00, 0x76bbff01, 0x80735725, 0xb5e0f137, 0x96abb417, 0xfb2cf5cf, 0x314cf55e, 0x77bde10e,
0x2ab24209, 0x81caa6F0, 0x3019b8eb, 0x427c505f, 0x5bba7d77, 0xf29cb4d6, 0x760f65ca, 0xf6b4536c,
0x190508ec, 0xc7764e22, 0x19b042a8, 0x2db4c5d8, 0x6e1cfd1d, 0x39bddd51, 0x942714ed, 0x19444d39,
0xed16e206, 0xc4102784, 0x590075fe, 0xaaef2ec6, 0xbeb84149, 0x8da31e4f, 0x7cbe7d77, 0x1c90b30a,
0xb5cd0704, 0x82144fd4, 0x4e6f4843, 0x106bc505, 0xf587fc97, 0x3665d9a3, 0x3ea014a8, 0xec664953,
0x6ec9e59e, 0xf9201e08, 0xf3676fb8, 0xe4e42c10, 0x92d33f64, 0x13b923f7, 0x308f7f50, 0xca98b420,
};
RNG rng(0);

2
modules/imgproc/test/test_contours.cpp
Unescape View File

@ -208,7 +208,7 @@ cvTsMarkContours( IplImage* img, int val )
int i, j;
int step = img->widthStep;
assert( img->depth == IPL_DEPTH_8U && img->nChannels == 1 && (val&1) != 0);
CV_Assert( img->depth == IPL_DEPTH_8U && img->nChannels == 1 && (val&1) != 0);
for( i = 1; i < img->height - 1; i++ )
for( j = 1; j < img->width - 1; j++ )

16
modules/imgproc/test/test_convhull.cpp
Unescape View File

@ -301,7 +301,7 @@ void CV_BaseShapeDescrTest::generate_point_set( void* pointsSet )
else
{
CvMat* ptm = (CvMat*)pointsSet;
assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
CV_Assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
total = ptm->rows + ptm->cols - 1;
point_type = CV_MAT_TYPE(ptm->type);
data = ptm->data.ptr;
@ -310,7 +310,7 @@ void CV_BaseShapeDescrTest::generate_point_set( void* pointsSet )
n = CV_MAT_CN(point_type);
point_type = CV_MAT_DEPTH(point_type);
assert( (point_type == CV_32S || point_type == CV_32F) && n <= 4 );
CV_Assert( (point_type == CV_32S || point_type == CV_32F) && n <= 4 );
for( i = 0; i < total; i++ )
{
@ -1335,7 +1335,7 @@ void CV_FitEllipseTest::generate_point_set( void* pointsSet )
else
{
CvMat* ptm = (CvMat*)pointsSet;
assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
CV_Assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
total = ptm->rows + ptm->cols - 1;
point_type = CV_MAT_TYPE(ptm->type);
data = ptm->data.ptr;
@ -1621,7 +1621,7 @@ void CV_FitLineTest::generate_point_set( void* pointsSet )
else
{
CvMat* ptm = (CvMat*)pointsSet;
assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
CV_Assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
total = ptm->rows + ptm->cols - 1;
point_type = CV_MAT_DEPTH(CV_MAT_TYPE(ptm->type));
data = ptm->data.ptr;
@ -1788,13 +1788,13 @@ cvTsGenerateTousledBlob( CvPoint2D32f center, CvSize2D32f axes,
else
{
CvMat* ptm = (CvMat*)points;
assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
CV_Assert( CV_IS_MAT(ptm) && CV_IS_MAT_CONT(ptm->type) );
total = ptm->rows + ptm->cols - 1;
point_type = CV_MAT_TYPE(ptm->type);
data = ptm->data.ptr;
}
assert( point_type == CV_32SC2 || point_type == CV_32FC2 );
CV_Assert( point_type == CV_32SC2 || point_type == CV_32FC2 );
for( i = 0; i < total; i++ )
{
@ -1874,8 +1874,8 @@ void CV_ContourMomentsTest::generate_point_set( void* pointsSet )
center.x = (float)(img_size.width*0.5 + (cvtest::randReal(rng)-0.5)*(img_size.width - max_sz*2)*0.8);
center.y = (float)(img_size.height*0.5 + (cvtest::randReal(rng)-0.5)*(img_size.height - max_sz*2)*0.8);
assert( 0 < center.x - max_sz && center.x + max_sz < img_size.width &&
0 < center.y - max_sz && center.y + max_sz < img_size.height );
CV_Assert( 0 < center.x - max_sz && center.x + max_sz < img_size.width &&
0 < center.y - max_sz && center.y + max_sz < img_size.height );
max_r_scale = cvtest::randReal(rng)*max_max_r_scale*0.01;
angle = cvtest::randReal(rng)*360;

2
modules/imgproc/test/test_distancetransform.cpp
Unescape View File

@ -161,7 +161,7 @@ cvTsDistTransform( const CvMat* _src, CvMat* _dst, int dist_type,
float delta[16];
int tstep, count;
assert( mask_size == 3 || mask_size == 5 );
CV_Assert( mask_size == 3 || mask_size == 5 );
if( dist_type == CV_DIST_USER )
memcpy( mask, _mask, sizeof(mask) );

8
modules/imgproc/test/test_filter.cpp
Unescape View File

@ -992,8 +992,8 @@ static void test_medianFilter( const Mat& src, Mat& dst, int m )
median_pair *buf0 = &_buf0[0], *buf1 = &_buf1[0];
int step = (int)(src.step/src.elemSize());
assert( src.rows == dst.rows + m - 1 && src.cols == dst.cols + m - 1 &&
src.type() == dst.type() && src.type() == CV_8UC1 );
CV_Assert( src.rows == dst.rows + m - 1 && src.cols == dst.cols + m - 1 &&
src.type() == dst.type() && src.type() == CV_8UC1 );
for( i = 0; i < dst.rows; i++ )
{
@ -1050,7 +1050,7 @@ static void test_medianFilter( const Mat& src, Mat& dst, int m )
*buf1++ = buf0[k++];
else
{
assert( col_buf[l] < INT_MAX );
CV_Assert( col_buf[l] < INT_MAX );
*buf1++ = median_pair(ins_col,col_buf[l++]);
}
}
@ -1061,7 +1061,7 @@ static void test_medianFilter( const Mat& src, Mat& dst, int m )
if( del_col < 0 )
n += m;
buf1 -= n;
assert( n == m2 );
CV_Assert( n == m2 );
dst1[j] = (uchar)buf1[n/2].val;
median_pair* tbuf;
CV_SWAP( buf0, buf1, tbuf );

6
modules/imgproc/test/test_imgwarp.cpp
Unescape View File

@ -169,7 +169,7 @@ int CV_ImgWarpBaseTest::prepare_test_case( int test_case_idx )
}
break;
default:
assert(0);
CV_Assert(0);
}
/*switch( depth )
@ -187,7 +187,7 @@ int CV_ImgWarpBaseTest::prepare_test_case( int test_case_idx )
((float*)ptr)[j] = (float)buffer[j];
break;
default:
assert(0);
CV_Assert(0);
}*/
cv::Mat src(1, cols*cn, CV_32F, &buffer[0]);
cv::Mat dst(1, cols*cn, depth, ptr);
@ -482,7 +482,7 @@ static void test_remap( const Mat& src, Mat& dst, const Mat& mapx, const Mat& ma
}
break;
default:
assert(0);
CV_Assert(0);
}
}
}

2
modules/imgproc/test/test_imgwarp_strict.cpp
Unescape View File

@ -935,7 +935,7 @@ void CV_Remap_Test::remap_generic(const Mat& _src, Mat& _dst)
else if (interpolation == INTER_LANCZOS4)
ksize = 8;
else if (interpolation != INTER_LINEAR)
assert(0);
CV_Assert(0);
int ofs = (ksize / 2) - 1;
CV_Assert(_src.depth() == CV_32F && _dst.type() == _src.type());

2
modules/imgproc/test/test_templmatch.cpp
Unescape View File

@ -185,7 +185,7 @@ static void cvTsMatchTemplate( const CvMat* img, const CvMat* templ, CvMat* resu
b_denom = 1.;
}
assert( CV_TM_SQDIFF <= method && method <= CV_TM_CCOEFF_NORMED );
CV_Assert( CV_TM_SQDIFF <= method && method <= CV_TM_CCOEFF_NORMED );
for( i = 0; i < result->rows; i++ )
{

2
modules/imgproc/test/test_thresh.cpp
Unescape View File

@ -474,7 +474,7 @@ static void test_threshold( const Mat& _src, Mat& _dst,
}
break;
default:
assert(0);
CV_Assert(0);
}
}

1
modules/ml/src/precomp.hpp
Unescape View File

@ -48,7 +48,6 @@
#include "opencv2/core/private.hpp"
#include <assert.h>
#include <float.h>
#include <limits.h>
#include <math.h>

2
modules/ml/src/tree.cpp
Unescape View File

@ -869,7 +869,7 @@ DTreesImpl::WSplit DTreesImpl::findSplitCatClass( int vi, const vector<int>& _si
}
else
{
assert( m == 2 );
CV_Assert( m == 2 );
dbl_ptr = (double**)(c_weights + _mi);
for( j = 0; j < mi; j++ )
dbl_ptr[j] = cjk + j*2 + 1;

8
modules/objdetect/src/cascadedetect.cpp
Unescape View File

@ -969,10 +969,10 @@ int CascadeClassifierImpl::runAt( Ptr<FeatureEvaluator>& evaluator, Point pt, in
{
CV_INSTRUMENT_REGION();
assert( !oldCascade &&
(data.featureType == FeatureEvaluator::HAAR ||
data.featureType == FeatureEvaluator::LBP ||
data.featureType == FeatureEvaluator::HOG) );
CV_Assert( !oldCascade &&
(data.featureType == FeatureEvaluator::HAAR ||
data.featureType == FeatureEvaluator::LBP ||
data.featureType == FeatureEvaluator::HOG) );
if( !evaluator->setWindow(pt, scaleIdx) )
return -1;

4
modules/objdetect/src/hog.cpp
Unescape View File

@ -854,7 +854,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
data->gradWeight = weights(i,j);
}
assert( count1 + count2 + count4 == rawBlockSize );
CV_Assert( count1 + count2 + count4 == rawBlockSize );
// defragment pixData
for( j = 0; j < count2; j++ )
pixData[j + count1] = pixData[j + rawBlockSize];
@ -876,7 +876,7 @@ void HOGCache::init(const HOGDescriptor* _descriptor,
const float* HOGCache::getBlock(Point pt, float* buf)
{
float* blockHist = buf;
assert(descriptor != 0);
CV_Assert(descriptor != 0);
// Size blockSize = descriptor->blockSize;
pt += imgoffset;

47
modules/objdetect/src/qrcode_encoder.cpp
Unescape View File

@ -308,18 +308,7 @@ int QRCodeEncoderImpl::versionAuto(const std::string& input_str)
for(size_t i = 0; i < possible_version.size(); i++)
{
int version_range_index = possible_version[i];
if (version_range_index == 1)
{
tmp_version = 1;
}
else if (version_range_index == 2)
{
tmp_version = 10;
}
else
{
tmp_version = 27;
}
encodeAuto(input_str, payload_tmp);
tmp_version = findVersionCapacity((int)payload_tmp.size(), ecc_level,
version_range[version_range_index], version_range[version_range_index + 1]);
@ -351,10 +340,11 @@ void QRCodeEncoderImpl::generateQR(const std::string &input)
int segment_begin = i * segment_len;
int segemnt_end = min((i + 1) * segment_len, (int) input.length()) - 1;
std::string input_info = input.substr(segment_begin, segemnt_end - segment_begin + 1);
int v = versionAuto(input_info);
int detected_version = versionAuto(input_info);
CV_Assert(detected_version != -1);
if (version_level == 0)
version_level = v;
else if (version_level < v)
version_level = detected_version;
else if (version_level < detected_version)
CV_Error(Error::StsBadArg, "The given version is not suitable for the given input string length ");
payload.clear();
@ -752,12 +742,14 @@ void QRCodeEncoderImpl::eccGenerate(vector<vector<uint8_t> > &data_blocks, vecto
void QRCodeEncoderImpl::rearrangeBlocks(const vector<vector<uint8_t> > &data_blocks, const vector<vector<uint8_t> > &ecc_blocks)
{
rearranged_data.clear();
rearranged_data.reserve(MAX_PAYLOAD_LEN);
int blocks = cur_ecc_params->num_blocks_in_G2 + cur_ecc_params->num_blocks_in_G1;
int col_border = max(cur_ecc_params->data_codewords_in_G2, cur_ecc_params->data_codewords_in_G1);
int total_codeword_num = version_info->total_codewords;
int is_not_equal = cur_ecc_params->data_codewords_in_G2 - cur_ecc_params->data_codewords_in_G1;
for (int i = 0; i < total_codeword_num; i++)
int add_steps = cur_ecc_params->data_codewords_in_G2 > cur_ecc_params->data_codewords_in_G1 ?
(cur_ecc_params->data_codewords_in_G2 - cur_ecc_params->data_codewords_in_G1) * cur_ecc_params->num_blocks_in_G1 : 0;
rearranged_data.reserve(total_codeword_num + add_steps);
for (int i = 0; i < total_codeword_num + add_steps; i++)
{
int cur_col = i / blocks;
int cur_row = i % blocks;
@ -783,16 +775,6 @@ void QRCodeEncoderImpl::rearrangeBlocks(const vector<vector<uint8_t> > &data_blo
}
rearranged_data.push_back(tmp);
}
const int remainder_len []= {0,
0, 7, 7, 7, 7, 7, 0, 0, 0, 0,
0, 0, 0, 3, 3, 3, 3, 3, 3, 3,
4, 4, 4, 4, 4, 4, 4, 3, 3, 3,
3, 3, 3, 3, 0, 0, 0, 0, 0, 0};
int cur_remainder_len = remainder_len[version_level];
if (cur_remainder_len != 0)
{
rearranged_data.push_back(0);
}
}
void QRCodeEncoderImpl::findAutoMaskType()
@ -1078,6 +1060,8 @@ void QRCodeEncoderImpl::writeData()
int dir = -1;
int count = 0;
int codeword_value = rearranged_data[0];
const int limit_bits = (int)rearranged_data.size() * 8;
bool limit_reached = false;
while (x > 0)
{
if (x == 6)
@ -1093,11 +1077,20 @@ void QRCodeEncoderImpl::writeData()
continue;
}
count++;
if (count == limit_bits)
{
limit_reached = true;
break;
}
if (count % 8 == 0)
{
codeword_value = rearranged_data[count / 8];
}
}
if (limit_reached)
{
break;
}
y += dir;
if (y < 0 || y >= version_size)
{

12
modules/objdetect/test/test_cascadeandhog.cpp
Unescape View File

@ -192,7 +192,7 @@ void CV_DetectorTest::run( int )
// write detectors
validationFS << DETECTORS << "{";
assert( detectorNames.size() == detectorFilenames.size() );
CV_Assert( detectorNames.size() == detectorFilenames.size() );
nit = detectorNames.begin();
for( int di = 0; nit != detectorNames.end(); ++nit, di++ )
{
@ -292,7 +292,7 @@ static bool isZero( uchar i ) {return i == 0;}
int CV_DetectorTest::validate( int detectorIdx, vector<vector<Rect> >& objects )
{
assert( imageFilenames.size() == objects.size() );
CV_Assert( imageFilenames.size() == objects.size() );
int imageIdx = 0;
int totalNoPair = 0, totalValRectCount = 0;
@ -489,7 +489,7 @@ int CV_HOGDetectorTest::detectMultiScale( int di, const Mat& img,
if( detectorFilenames[di].empty() )
hog.setSVMDetector(HOGDescriptor::getDefaultPeopleDetector());
else
assert(0);
CV_Assert(0);
hog.detectMultiScale(img, objects);
return cvtest::TS::OK;
}
@ -772,7 +772,7 @@ void HOGCacheTester::init(const HOGDescriptorTester* _descriptor,
data->gradWeight = weights(i,j);
}
assert( count1 + count2 + count4 == rawBlockSize );
CV_Assert( count1 + count2 + count4 == rawBlockSize );
// defragment pixData
for( j = 0; j < count2; j++ )
pixData[j + count1] = pixData[j + rawBlockSize];
@ -794,7 +794,7 @@ void HOGCacheTester::init(const HOGDescriptorTester* _descriptor,
const float* HOGCacheTester::getBlock(Point pt, float* buf)
{
float* blockHist = buf;
assert(descriptor != 0);
CV_Assert(descriptor != 0);
Size blockSize = descriptor->blockSize;
pt += imgoffset;
@ -1275,7 +1275,7 @@ void HOGDescriptorTester::computeGradient(InputArray _img, InputOutputArray _gra
hidx += _nbins;
else if( hidx >= _nbins )
hidx -= _nbins;
assert( (unsigned)hidx < (unsigned)_nbins );
CV_Assert( (unsigned)hidx < (unsigned)_nbins );
qanglePtr[x*2] = (uchar)hidx;
hidx++;

2
modules/ts/src/ocl_perf.cpp
Unescape View File

@ -50,7 +50,7 @@ namespace perf {
void checkDeviceMaxMemoryAllocSize(const Size& size, int type, int factor)
{
assert(factor > 0);
CV_Assert(factor > 0);
if (!cv::ocl::useOpenCL())
return;

2
modules/ts/src/ts.cpp
Unescape View File

@ -379,7 +379,7 @@ void BaseTest::run( int start_from )
void BaseTest::run_func(void)
{
assert(0);
CV_Assert(0);
}

6
modules/ts/src/ts_arrtest.cpp
Unescape View File

@ -268,14 +268,14 @@ void ArrayTest::fill_array( int /*test_case_idx*/, int i, int j, Mat& arr )
double ArrayTest::get_success_error_level( int /*test_case_idx*/, int i, int j )
{
int elem_depth = CV_MAT_DEPTH(cvGetElemType(test_array[i][j]));
assert( i == OUTPUT || i == INPUT_OUTPUT );
CV_Assert( i == OUTPUT || i == INPUT_OUTPUT );
return elem_depth < CV_32F ? 0 : elem_depth == CV_32F ? FLT_EPSILON*100: DBL_EPSILON*5000;
}
void ArrayTest::prepare_to_validation( int /*test_case_idx*/ )
{
assert(0);
CV_Assert(0);
}
@ -293,7 +293,7 @@ int ArrayTest::validate_test_results( int test_case_idx )
int i1 = i == 0 ? REF_OUTPUT : REF_INPUT_OUTPUT;
size_t sizei = test_array[i0].size();
assert( sizei == test_array[i1].size() );
CV_Assert( sizei == test_array[i1].size() );
for( j = 0; j < sizei; j++ )
{
double err_level;

8
modules/ts/src/ts_func.cpp
Unescape View File

@ -2122,7 +2122,7 @@ int cmpEps( const Mat& arr_, const Mat& refarr_, double* _realmaxdiff,
}
break;
default:
assert(0);
CV_Assert(0);
return CMP_EPS_BIG_DIFF;
}
if(_realmaxdiff)
@ -2733,7 +2733,7 @@ static void calcSobelKernel1D( int order, int _aperture_size, int size, vector<i
if( _aperture_size < 0 )
{
static const int scharr[8] = { 3, 10, 3, -1, 0, 1, 0, 0 }; // extra elements to eliminate "-Warray-bounds" bogus warning
assert( size == 3 );
CV_Assert( size == 3 && order < 2 );
for( i = 0; i < size; i++ )
kernel[i] = scharr[order*3 + i];
return;
@ -3057,7 +3057,7 @@ void threshold( const Mat& _src, Mat& _dst,
imaxval = cvRound(maxval);
}
assert( depth == CV_8U || depth == CV_16S || depth == CV_32F );
CV_Assert( depth == CV_8U || depth == CV_16S || depth == CV_32F );
switch( thresh_type )
{
@ -3219,7 +3219,7 @@ void threshold( const Mat& _src, Mat& _dst,
}
break;
default:
assert(0);
CV_Assert(0);
}
}

8
modules/ts/src/ts_perf.cpp
Unescape View File

@ -1366,7 +1366,7 @@ bool TestBase::next()
bool has_next = false;
do {
assert(currentIter == times.size());
CV_Assert(currentIter == times.size());
if (currentIter == 0)
{
has_next = true;
@ -1379,7 +1379,7 @@ bool TestBase::next()
}
else
{
assert(getCurrentPerformanceStrategy() == PERF_STRATEGY_SIMPLE);
CV_Assert(getCurrentPerformanceStrategy() == PERF_STRATEGY_SIMPLE);
if (totalTime - lastActivityPrintTime >= cv::getTickFrequency() * 10)
{
std::cout << '.' << std::endl;
@ -1638,7 +1638,7 @@ performance_metrics& TestBase::calcMetrics()
}
else
{
assert(false);
CV_Assert(false);
}
int offset = static_cast<int>(start - times.begin());
@ -1714,7 +1714,7 @@ void TestBase::validateMetrics()
}
else
{
assert(false);
CV_Assert(false);
}
}

9
modules/videoio/src/cap_ffmpeg_impl.hpp
Unescape View File

@ -47,7 +47,6 @@ using namespace cv;
#if !(defined(_WIN32) || defined(WINCE))
# include <pthread.h>
#endif
#include <assert.h>
#include <algorithm>
#include <limits>
@ -400,7 +399,7 @@ static
inline int _opencv_ffmpeg_interrupt_callback(void *ptr)
{
AVInterruptCallbackMetadata* metadata = (AVInterruptCallbackMetadata*)ptr;
assert(metadata);
CV_Assert(metadata);
if (metadata->timeout_after_ms == 0)
{
@ -2165,7 +2164,7 @@ bool CvVideoWriter_FFMPEG::writeFrame( const unsigned char* data, int step, int
}
}
else {
assert(false);
CV_Assert(false);
}
if( (width & -2) != frame_width || (height & -2) != frame_height || !data )
@ -2218,7 +2217,7 @@ bool CvVideoWriter_FFMPEG::writeFrame( const unsigned char* data, int step, int
sw_pix_fmt = ((AVHWFramesContext*)c->hw_frames_ctx->data)->sw_format;
#endif
if ( sw_pix_fmt != input_pix_fmt ) {
assert( input_picture );
CV_Assert( input_picture );
// let input_picture point to the raw data buffer of 'image'
_opencv_ffmpeg_av_image_fill_arrays(input_picture, (uint8_t *) data,
(AVPixelFormat)input_pix_fmt, width, height);
@ -2592,7 +2591,7 @@ bool CvVideoWriter_FFMPEG::open( const char * filename, int fourcc,
// alloc memory for context
oc = avformat_alloc_context();
assert (oc);
CV_Assert(oc);
/* set file name */
oc->oformat = fmt;

5
modules/videoio/src/cap_v4l.cpp
Unescape View File

@ -220,7 +220,6 @@ make & enjoy!
#include <string.h>
#include <stdlib.h>
#include <assert.h>
#include <sys/stat.h>
#include <sys/ioctl.h>
#include <limits>
@ -938,8 +937,8 @@ bool CvCaptureCAM_V4L::read_frame_v4l2()
return false;
}
assert(buf.index < req.count);
assert(buffers[buf.index].length == buf.length);
CV_Assert(buf.index < req.count);
CV_Assert(buffers[buf.index].length == buf.length);
//We shouldn't use this buffer in the queue while not retrieve frame from it.
buffers[buf.index].buffer = buf;

1
modules/videoio/src/precomp.hpp
Unescape View File

@ -82,7 +82,6 @@
#include <string.h>
#include <limits.h>
#include <ctype.h>
#include <assert.h> // FIXIT remove this
#if defined _WIN32 || defined WINCE
#if !defined _WIN32_WINNT

Write Preview
Loading…
Cancel
Save