Commit Graph

238 Commits

Author SHA1 Message Date
peng xiao
dcc4766129 Fix white-spacing 2013-06-27 09:57:42 +08:00
peng xiao
9270205947 Fix build errors 2013-06-26 17:24:09 +08:00
peng xiao
132b885b24 Add opencl implementation of Farnback optical flow. 2013-06-26 16:35:19 +08:00
Roman Donchenko
171a56fe82 Merge pull request #1034 from pengx17:2.4_oclmat_inoutarray 2013-06-24 12:27:26 +04:00
Roman Donchenko
7cb047e2d9 Merge pull request #1031 from pengx17:2.4_fix_converTo 2013-06-24 12:25:48 +04:00
peng xiao
290c8db0a8 Revise naming for getOclMat function. 2013-06-21 14:51:23 +08:00
peng xiao
e129638263 Add a workaround to interpolate between oclMat and Input/OutputArray. 2013-06-21 14:05:29 +08:00
peng xiao
c1f4fe1637 Fix a bug of convertTo.
The bug was found that all 3-channel oclMat's were converted to
4-channel oclMat's after using convertTo function.
2013-06-20 11:26:22 +08:00
yao
26c246140a optimize hog 2013-06-19 11:20:45 +08:00
Roman Donchenko
47ea0614dc Merge pull request #969 from pengx17:2.4_binary_cache 2013-06-14 15:47:23 +04:00
peng xiao
e6b18fc492 Fix a bug caused by NDEBUG macro; it is now removed.
Revise some descriptions of the enums.
2013-06-14 16:37:00 +08:00
Roman Donchenko
b725cbf73f Merge pull request #986 from pengx17:2.4_initiated_context 2013-06-13 16:55:56 +04:00
peng xiao
5fd724b54a Add a function to query if global OpenCL context is initialized. 2013-06-13 10:46:12 +08:00
Peng Xiao
d9ab22e4ed Fix two bugs related to opencl context.
1. As getDevice will implicitly call setDevice, in getContext we should not need to call it again.
2. Fix an incorrect type casting.
2013-06-12 13:55:20 +08:00
Peng Xiao
1d8cd3a717 Add ocl CLACH implementation.
Test cases (accuracy and performance) are provided.
2013-06-10 18:37:48 +08:00
Peng Xiao
e77abeef16 Add a new global function to control ocl binary storage
Previously the feature is controlled by setBinpath implicitly.
We add the function to cope with setBinpath and setBinpath is only
useful when setBinaryDiskCache is set.
Refer to the header to see more info.
2013-06-10 16:38:22 +08:00
peng xiao
429f84e59e Fix a bug of cornerHarris and cornerMinEigenVal.
The bug is a buffer overrun when border type is reflect101.
It is found that gfft crashed with input of size 100x100 on Intel CPU.
2013-06-06 11:44:35 +08:00
Vadim Pisarevsky
119b7a298d Merge pull request #933 from pengx17:2.4_macfix_cont 2013-06-04 13:34:53 +04:00
Vadim Pisarevsky
2d88f20c1e Merge pull request #935 from pengx17:2.4_filter2d_fix 2013-06-04 13:34:40 +04:00
Vadim Pisarevsky
e0c2daadbf Merge pull request #937 from bitwangyaoyao:2.4_fixPyrLK 2013-06-04 13:34:09 +04:00
peng xiao
a7a94de74a Fix a bug of gfft.
When user provided corners buffer is big enough to be copied to from
tmpCorners_, we allow the buffer to be reused other than allocate a new
cl_mem object.
2013-06-04 15:55:33 +08:00
peng xiao
97b86aa259 Initialize OpenCL context at the end of getDevice call.
Added for better compatibility with the current samples/test cases.
User now will be able to initialize OpenCL context explicitly with
ocl::getDevice api.
This may be obsoleted in future releases.
2013-05-31 16:48:40 +08:00
yao
15a213d3fc fix a crash on Linux 2013-05-31 15:35:54 +08:00
peng xiao
b1c248fcc9 Fix ocl::filter2D.
In current implementation, this function only works when anchor point is
in the kernel center and kernel size supported is either 3x3 or 5x5.
2013-05-31 10:53:52 +08:00
peng xiao
fdc133d8c9 Fix ocl::pyrup kernel build on Mac. 2013-05-30 16:34:20 +08:00
Vadim Pisarevsky
37091b086c Merge pull request #932 from bitwangyaoyao:2.4_fixBFM 2013-05-30 12:21:27 +04:00
Vadim Pisarevsky
5a4efe8bcf Merge pull request #924 from pengx17:2.4_arithm_fix 2013-05-30 12:01:25 +04:00
Vadim Pisarevsky
678371be39 Merge pull request #925 from pengx17:2.4_canny_tmp_fix 2013-05-30 12:01:09 +04:00
yao
7ed9c0e87a Fix brute_force_matcher's hung on some Intel CPU OCL 2013-05-30 14:57:15 +08:00
peng xiao
fd7ba355ee Add non-stump based ocl Haar cascade classifier support.
For example, haarcascade_frontalface_alt2.xml is now supported.
Note that classifier's pattern of a cascade file must be consistent,
i.e., all trees must either have two nodes or one node, otherwise
unexpected results will occur.

Other fixes:
Test cases are updated.
Some unused codes are removed.
Fix some problems of haar when using OclCascadeClassifierBuf.
2013-05-30 14:01:19 +08:00
peng xiao
006e4242b2 Merge branch '2.4' of https://github.com/Itseez/opencv into 2.4_oclgfft 2013-05-29 17:57:14 +08:00
peng xiao
d015fa76fa Fix 2.4 ocl Canny.
This fix is a workaround for current 2.4 branch without introducing an
additional oclMat buffer into CannyBuf object.
Test case is cleaned up.
Volatile keywords in kernels are removed for performance concern.
2013-05-29 14:15:26 +08:00
Vadim Pisarevsky
2ccdf56119 Merge pull request #913 from janm399:2.4 2013-05-28 22:05:14 +04:00
Vadim Pisarevsky
24ca620c32 Merge pull request #916 from bitwangyaoyao:2.4_fixPyrLK 2013-05-28 22:04:37 +04:00
yao
d81c145fa9 fix memory leak 2013-05-28 18:07:41 +08:00
peng xiao
1d0c283508 Fix a bug when pushing pointers of arguments into std::vector.
When argument pointers pushed into an vector and the pointers point to
address on stack, we need to make sure they are valid until kernels are
successfully flushed onto the queue.
2013-05-28 17:27:55 +08:00
peng xiao
6fae02c05d Fix some OpenCL kernel file build errors on Mac. 2013-05-28 11:12:05 +08:00
yao
c58e0d5d73 fix hog 2013-05-24 13:46:21 +08:00
yao
d45f9ef866 fix Linux build errors 2013-05-23 17:58:50 +08:00
yao
a223b5624f fix pyrLK's mismatch on Intel GPUs 2013-05-23 10:55:08 +08:00
Jan Machacek
46b770f255 Fixed include name in OpenCL on OS X 2013-05-22 13:22:16 +01:00
Roman Donchenko
519ee72f6f Merge pull request #905 from bitwangyaoyao:2.4_TVL1 2013-05-22 12:08:16 +04:00
peng xiao
b4a4a05bdc Add ocl's good features to track implementation.
Additional notes with this commit:
1. Add cornerHarris_dxdy and cornerMinEigenVal_dxdy to get
the interim dx and dy output of Sobel operator;
2. Add minMax_buf to allow user to reuse buffers in minMax;
3. Fix an error when either min or max pointer fed into minMax is NULL;
4. Corner sorter temporarily uses C++ STL's quick sort. A parallel
 selection sort in OpneCL is contained in the implementation but disabled
due to poor performance at the moment.
5. Accuracy test for ocl gfft.
2013-05-22 13:46:42 +08:00
Vadim Pisarevsky
cbbc82a789 Merge pull request #882 from pengx17:2.4_rewrite_query_info 2013-05-20 14:01:03 +04:00
yao
d8b192c84d Fix the mismatch on NV GPUs 2013-05-20 14:46:17 +08:00
yao
4162ebfad3 add OpticalFlowDual_TVL1_OCL function 2013-05-17 15:34:22 +08:00
peng xiao
1ecc765903 Merge branch '2.4' of https://github.com/Itseez/opencv into 2.4_pyrup_fix 2013-05-15 08:55:43 +08:00
peng xiao
df3997b108 Fix ocl::pyrUp
Use predefined OpenCL function to convert integers to floating points.
This is more accurate than before as it enables:
1. saturate cast
2. customized rounding
2013-05-15 08:51:21 +08:00
Vadim Pisarevsky
25e6902a43 Merge pull request #818 from bitwangyaoyao:2.4_optBlur 2013-05-14 15:56:56 +04:00
Vadim Pisarevsky
bd1d7cd208 Merge pull request #839 from pengx17:2.4_ocl_csbp 2013-05-14 15:56:07 +04:00