platform/upstream/opencv.git
7 years agoMerge pull request #9826 from p0wdrdotcom:master
Alexander Alekhin [Wed, 11 Oct 2017 08:55:15 +0000 (08:55 +0000)]
Merge pull request #9826 from p0wdrdotcom:master

7 years agoThis statement was keeping HAAR cascades from leveraging opencl on nvidia devices...
Geoff McIver [Tue, 10 Oct 2017 20:32:38 +0000 (09:32 +1300)]
This statement was keeping HAAR cascades from leveraging opencl on nvidia devices. "localSize" on the featureEvaluator remains Size(0, 0) which sets the bool "use_ocl" to false. Adding this allows NVidia GPUs to leverage opencl HAAR Cascades

7 years agoMerge pull request #9816 from opalmirror:fix_stereobm_mindisp_truncation_1
Vadim Pisarevsky [Tue, 10 Oct 2017 14:23:02 +0000 (14:23 +0000)]
Merge pull request #9816 from opalmirror:fix_stereobm_mindisp_truncation_1

7 years agoMerge pull request #9822 from alalek:fix_tbb_search
Vadim Pisarevsky [Tue, 10 Oct 2017 14:22:00 +0000 (14:22 +0000)]
Merge pull request #9822 from alalek:fix_tbb_search

7 years agoMerge pull request #8940 from 678098:nonblocking_haar_detector_parallel_for
Vadim Pisarevsky [Tue, 10 Oct 2017 13:51:39 +0000 (13:51 +0000)]
Merge pull request #8940 from 678098:nonblocking_haar_detector_parallel_for

7 years agoMerge pull request #9705 from AlexeyAB:dnn_darknet_yolo_v2
Vadim Pisarevsky [Tue, 10 Oct 2017 12:02:03 +0000 (12:02 +0000)]
Merge pull request #9705 from AlexeyAB:dnn_darknet_yolo_v2

7 years agotbb: don't search library in ENV{LD_LIBRARY_PATH}
Alexander Alekhin [Tue, 10 Oct 2017 11:18:28 +0000 (14:18 +0300)]
tbb: don't search library in ENV{LD_LIBRARY_PATH}

ENV{LIBRARY_PATH} is enough

7 years agoMerge pull request #9786 from LaurentBerger:Histo3d
Vadim Pisarevsky [Tue, 10 Oct 2017 10:58:34 +0000 (10:58 +0000)]
Merge pull request #9786 from LaurentBerger:Histo3d

7 years agoMerge pull request #9750 from dkurt:feature_dnn_tf_text_graph
Vadim Pisarevsky [Tue, 10 Oct 2017 10:06:24 +0000 (10:06 +0000)]
Merge pull request #9750 from dkurt:feature_dnn_tf_text_graph

7 years agoMerge pull request #9763 from seiko2plus:addVsxCore
Vadim Pisarevsky [Tue, 10 Oct 2017 10:00:31 +0000 (10:00 +0000)]
Merge pull request #9763 from seiko2plus:addVsxCore

7 years agoMerge pull request #9811 from dkurt:prelu_with_shared_channels
Vadim Pisarevsky [Tue, 10 Oct 2017 09:57:51 +0000 (09:57 +0000)]
Merge pull request #9811 from dkurt:prelu_with_shared_channels

7 years agoMerge pull request #9805 from alalek:ffmpeg_update
Maksim Shabunin [Tue, 10 Oct 2017 09:02:02 +0000 (09:02 +0000)]
Merge pull request #9805 from alalek:ffmpeg_update

7 years agofix StereoBM disparity map right margin truncation when minDisparities > 0
James Perkins [Thu, 5 Oct 2017 22:42:20 +0000 (15:42 -0700)]
fix StereoBM disparity map right margin truncation when minDisparities > 0

7 years agoIt's done
LaurentBerger [Mon, 9 Oct 2017 20:25:57 +0000 (22:25 +0200)]
It's done

7 years agoMerge pull request #9815 from alalek:issue_9813
Alexander Alekhin [Mon, 9 Oct 2017 20:08:00 +0000 (20:08 +0000)]
Merge pull request #9815 from alalek:issue_9813

7 years agoqt: fix non-world builds
Alexander Alekhin [Mon, 9 Oct 2017 19:12:52 +0000 (22:12 +0300)]
qt: fix non-world builds

7 years agoAdded DNN Darknet Yolo v2 for object detection
AlexeyAB [Sun, 24 Sep 2017 20:34:08 +0000 (23:34 +0300)]
Added DNN Darknet Yolo v2 for object detection

7 years agoPReLU layer from Caffe
Dmitry Kurtaev [Mon, 9 Oct 2017 14:47:46 +0000 (17:47 +0300)]
PReLU layer from Caffe

7 years agoMerge pull request #9810 from alalek:fix_js_build
Alexander Alekhin [Mon, 9 Oct 2017 16:19:09 +0000 (16:19 +0000)]
Merge pull request #9810 from alalek:fix_js_build

7 years agojs: skip 'utils' headers
Alexander Alekhin [Mon, 9 Oct 2017 13:14:01 +0000 (16:14 +0300)]
js: skip 'utils' headers

7 years agoMerge pull request #9792 from alalek:port_9776
Vadim Pisarevsky [Mon, 9 Oct 2017 12:46:06 +0000 (12:46 +0000)]
Merge pull request #9792 from alalek:port_9776

7 years agoMerge pull request #9799 from alalek:ocl_program
Vadim Pisarevsky [Mon, 9 Oct 2017 12:43:46 +0000 (12:43 +0000)]
Merge pull request #9799 from alalek:ocl_program

7 years agoMerge pull request #9800 from alalek:fix_build_msvs2010
Vadim Pisarevsky [Mon, 9 Oct 2017 12:33:08 +0000 (12:33 +0000)]
Merge pull request #9800 from alalek:fix_build_msvs2010

7 years agoMerge pull request #9803 from wzw-intel:ocl_timer
Vadim Pisarevsky [Mon, 9 Oct 2017 12:11:22 +0000 (12:11 +0000)]
Merge pull request #9803 from wzw-intel:ocl_timer

7 years agoMerge pull request #9806 from mshabunin:fix-static-7
Vadim Pisarevsky [Mon, 9 Oct 2017 12:09:55 +0000 (12:09 +0000)]
Merge pull request #9806 from mshabunin:fix-static-7

7 years agoFixed and suppressed warnings produced by GCC 6.3.0
Maksim Shabunin [Mon, 9 Oct 2017 10:52:07 +0000 (13:52 +0300)]
Fixed and suppressed warnings produced by GCC 6.3.0

7 years agoMerge pull request #9790 from mevatron:patch-1
Alexander Alekhin [Mon, 9 Oct 2017 10:46:50 +0000 (10:46 +0000)]
Merge pull request #9790 from mevatron:patch-1

7 years agoffmpeg: update 3.3.4
Alexander Alekhin [Mon, 9 Oct 2017 10:17:17 +0000 (13:17 +0300)]
ffmpeg: update 3.3.4

7 years agoFixed warnings produced by static analyzer
Maksim Shabunin [Mon, 9 Oct 2017 10:33:55 +0000 (13:33 +0300)]
Fixed warnings produced by static analyzer

7 years agoUse @snippet
LaurentBerger [Mon, 9 Oct 2017 08:59:30 +0000 (10:59 +0200)]
Use @snippet

7 years agoocl: simplify ocl::Timer
Wu Zhiwen [Mon, 2 Oct 2017 11:22:28 +0000 (19:22 +0800)]
ocl: simplify ocl::Timer

Use clFinish to gurantee commands completed, instead of waiting for events.

Signed-off-by: Wu Zhiwen <zhiwen.wu@intel.com>
7 years agoAdded universal intrinsic for VSX
Sayed Adel [Mon, 2 Oct 2017 22:55:59 +0000 (22:55 +0000)]
Added universal intrinsic for VSX

7 years agoAdded support for VSX
Sayed Adel [Mon, 2 Oct 2017 22:54:31 +0000 (22:54 +0000)]
Added support for VSX

7 years agobuild: fix MSVS2010
Alexander Alekhin [Sun, 8 Oct 2017 20:32:22 +0000 (23:32 +0300)]
build: fix MSVS2010

7 years agoText TensorFlow graphs parsing. MobileNet-SSD for 90 classes.
Dmitry Kurtaev [Thu, 28 Sep 2017 13:51:47 +0000 (16:51 +0300)]
Text TensorFlow graphs parsing. MobileNet-SSD for 90 classes.

7 years agoMerge pull request #9796 from lzx1413:patch-1
Alexander Alekhin [Sun, 8 Oct 2017 17:32:49 +0000 (17:32 +0000)]
Merge pull request #9796 from lzx1413:patch-1

7 years agoocl: refactor program compilation
Alexander Alekhin [Thu, 7 Sep 2017 13:41:19 +0000 (16:41 +0300)]
ocl: refactor program compilation

7 years agoocl: define ProgramSource before program
Alexander Alekhin [Thu, 7 Sep 2017 13:40:03 +0000 (16:40 +0300)]
ocl: define ProgramSource before program

no changes in code

7 years agodnn(samples): fix python syntax (false -> False)
Alexander Alekhin [Sun, 8 Oct 2017 16:17:40 +0000 (19:17 +0300)]
dnn(samples): fix python syntax (false -> False)

7 years agoin python, false should be False
lzx1413 [Sun, 8 Oct 2017 13:10:25 +0000 (21:10 +0800)]
in python, false should be False

7 years agoMerge pull request #9795 from IgWod:refactor-scalar-to-raw-data
Alexander Alekhin [Sun, 8 Oct 2017 09:48:46 +0000 (09:48 +0000)]
Merge pull request #9795 from IgWod:refactor-scalar-to-raw-data

7 years agoExtract code from scalarToRawData
Igor Wodiany [Sat, 7 Oct 2017 18:46:45 +0000 (19:46 +0100)]
Extract code from scalarToRawData

The same code was repeated several time for different data types, so
it was extracted as a templated function to improve maintability and
make a code more clear.

7 years agoFix a memory leak in the Mat copying constructor
Igor Wodiany [Wed, 4 Oct 2017 21:54:40 +0000 (22:54 +0100)]
Fix a memory leak in the Mat copying constructor

Exception may be rasied inside the body of a copying constructor after
refcount has been increased, and beacause in the case of the exception
destrcutor is never called what causes memory leak. This commit adds a
workaround that calls the release() function before the exception is
thrown outside the contructor.

7 years agoSmall fix for using Qt 5.x with opencv_world
mevatron [Fri, 6 Oct 2017 21:03:39 +0000 (16:03 -0500)]
Small fix for using Qt 5.x with opencv_world

7 years ago3D histogram
LaurentBerger [Thu, 5 Oct 2017 08:06:40 +0000 (10:06 +0200)]
3D histogram

7 years agoMerge pull request #9761 from Jazmann:ellipseFitAMS&Direct
Vadim Pisarevsky [Fri, 6 Oct 2017 14:35:42 +0000 (14:35 +0000)]
Merge pull request #9761 from Jazmann:ellipseFitAMS&Direct

7 years agoMerge pull request #9424 from Cartucho:update_imgproc_tutorials
Vadim Pisarevsky [Fri, 6 Oct 2017 14:02:58 +0000 (14:02 +0000)]
Merge pull request #9424 from Cartucho:update_imgproc_tutorials

7 years agoMerge pull request #9772 from dkurt:fix_caffe_eltwise_and_fc_layers
Vadim Pisarevsky [Fri, 6 Oct 2017 13:47:54 +0000 (13:47 +0000)]
Merge pull request #9772 from dkurt:fix_caffe_eltwise_and_fc_layers

7 years agoMerge pull request #9787 from dkurt:feature_dnn_resize_nearest_neighbor
Vadim Pisarevsky [Fri, 6 Oct 2017 13:46:50 +0000 (13:46 +0000)]
Merge pull request #9787 from dkurt:feature_dnn_resize_nearest_neighbor

7 years agoMerge pull request #9778 from dkurt:dnn_colorization
Vadim Pisarevsky [Fri, 6 Oct 2017 11:48:05 +0000 (11:48 +0000)]
Merge pull request #9778 from dkurt:dnn_colorization

7 years agoMerge pull request #9779 from Lightricks:feature/assetslibrary-remove-link
Vadim Pisarevsky [Fri, 6 Oct 2017 11:45:35 +0000 (11:45 +0000)]
Merge pull request #9779 from Lightricks:feature/assetslibrary-remove-link

7 years agoNearest neighbor resize layer
Dmitry Kurtaev [Fri, 6 Oct 2017 11:24:01 +0000 (14:24 +0300)]
Nearest neighbor resize layer

7 years agoMerge pull request #9773 from shoeffner:feature/doxygen-tag-file
Maksim Shabunin [Fri, 6 Oct 2017 10:44:58 +0000 (10:44 +0000)]
Merge pull request #9773 from shoeffner:feature/doxygen-tag-file

7 years agoGrayscale colorization model (https://github.com/richzhang/colorization) test.
Dmitry Kurtaev [Thu, 5 Oct 2017 10:04:22 +0000 (13:04 +0300)]
Grayscale colorization model (https://github.com/richzhang/colorization) test.

7 years agovideoio: remove AssetsLibrary dependency.
Yaron Inger [Thu, 5 Oct 2017 17:22:56 +0000 (20:22 +0300)]
videoio: remove AssetsLibrary dependency.

The entire AssetsLibrary framework is deprecated since iOS 8.0. The code
used in the camera example code can use UIKit to save videos to the
camera instead, which allows to avoid linking with PhotoKit instead to
prevent increasing the iOS deployment target.

7 years agoTutorial Hough Lines
tribta [Sat, 26 Aug 2017 23:02:29 +0000 (00:02 +0100)]
Tutorial Hough Lines

7 years agoTutorial Hough Circles
tribta [Sat, 26 Aug 2017 14:11:46 +0000 (15:11 +0100)]
Tutorial Hough Circles

7 years agoTutorial Make Border
tribta [Fri, 25 Aug 2017 18:44:55 +0000 (19:44 +0100)]
Tutorial Make Border

7 years agoTutorial Filter2D
tribta [Thu, 24 Aug 2017 15:22:27 +0000 (16:22 +0100)]
Tutorial Filter2D

7 years agoTutorial Image Pyramids
tribta [Wed, 23 Aug 2017 14:37:11 +0000 (15:37 +0100)]
Tutorial Image Pyramids

7 years agoTutorial Sobel Derivatives
tribta [Tue, 22 Aug 2017 11:28:23 +0000 (12:28 +0100)]
Tutorial Sobel Derivatives

7 years agoTutorial Laplace Operator
tribta [Mon, 21 Aug 2017 23:17:09 +0000 (00:17 +0100)]
Tutorial Laplace Operator

7 years agoTutorial Morph Lines Detection
tribta [Mon, 21 Aug 2017 17:51:50 +0000 (18:51 +0100)]
Tutorial Morph Lines Detection

7 years agoTutorial Hit-or-Miss
tribta [Sun, 20 Aug 2017 21:54:38 +0000 (22:54 +0100)]
Tutorial Hit-or-Miss

7 years agoTutorial Smoothing Images
tribta [Sun, 20 Aug 2017 16:14:12 +0000 (17:14 +0100)]
Tutorial Smoothing Images

7 years agoMerge pull request #9762 from dkurt:fix_tensorflow_split_layer
Vadim Pisarevsky [Thu, 5 Oct 2017 10:51:49 +0000 (10:51 +0000)]
Merge pull request #9762 from dkurt:fix_tensorflow_split_layer

7 years agoMerge pull request #9771 from Bisaloo:patch-1
Vadim Pisarevsky [Thu, 5 Oct 2017 10:48:20 +0000 (10:48 +0000)]
Merge pull request #9771 from Bisaloo:patch-1

7 years agoConfiguring Doxygen to generate a tag file.
Sebastian Höffner [Wed, 4 Oct 2017 13:52:12 +0000 (15:52 +0200)]
Configuring Doxygen to generate a tag file.

Additionally adds a tutorial on how to use the tag file in other
projects to link back to the OpenCV documentation.

The tutorial is also the only place which links to the tag file as of
now.

For more information on tag files and external documentation, check
https://www.stack.nl/~dimitri/doxygen/manual/external.html

7 years agoMerge pull request #9769 from pstieber:FixPrivateIncludeInSaturateCast
Alexander Alekhin [Wed, 4 Oct 2017 12:49:44 +0000 (12:49 +0000)]
Merge pull request #9769 from pstieber:FixPrivateIncludeInSaturateCast

7 years agoMultidimensional eltwise layer.
Dmitry Kurtaev [Wed, 4 Oct 2017 09:23:35 +0000 (12:23 +0300)]
Multidimensional eltwise layer.
Fixed fully-connected layer axis.

7 years agoFix 404 url in canny edge detection tutorial
Bisaloo [Wed, 4 Oct 2017 09:16:06 +0000 (11:16 +0200)]
Fix 404 url in canny edge detection tutorial

7 years agoReplace private.cuda.hpp with conditional include of cuda_fp16.h.
Peter J. Stieber [Wed, 4 Oct 2017 00:47:52 +0000 (17:47 -0700)]
Replace private.cuda.hpp with conditional include of cuda_fp16.h.

7 years agoFix TensorFlow split layer
Dmitry Kurtaev [Mon, 2 Oct 2017 19:44:42 +0000 (22:44 +0300)]
Fix TensorFlow split layer

7 years agoFix Transposed eigenvals and vecs. Didn't notice at first
Jasper Shemilt [Mon, 2 Oct 2017 16:56:08 +0000 (17:56 +0100)]
Fix Transposed eigenvals and vecs. Didn't notice at first

7 years agoAdds fitEllipseAMS to imgproc: The Approximate Mean Square (AMS) proposed by Taubin...
Jasper Shemilt [Mon, 2 Oct 2017 15:38:41 +0000 (16:38 +0100)]
Adds fitEllipseAMS to imgproc: The Approximate Mean Square (AMS) proposed by Taubin 1991.
Adds fitEllipseDirect to imgproc: The Direct least square (Direct) method by Fitzgibbon1999.

New Tests are included for the methods.
fitEllipseAMS Tests
fitEllipseDirect Tests

Comparative examples are added to fitEllipse.cpp in Samples.

7 years agoMerge pull request #9754 from sturkmen72:patch-1
Alexander Alekhin [Mon, 2 Oct 2017 15:26:37 +0000 (15:26 +0000)]
Merge pull request #9754 from sturkmen72:patch-1

7 years agoMerge pull request #9744 from sturkmen72:Update_HOGDescriptor_Documentation
Vadim Pisarevsky [Mon, 2 Oct 2017 14:50:46 +0000 (14:50 +0000)]
Merge pull request #9744 from sturkmen72:Update_HOGDescriptor_Documentation

7 years agoMerge pull request #9596 from paroj:circle_board_params
Alexander Alekhin [Mon, 2 Oct 2017 14:47:21 +0000 (14:47 +0000)]
Merge pull request #9596 from paroj:circle_board_params

7 years agoMerge pull request #9114 from pengli:dnn_rebase
pengli [Mon, 2 Oct 2017 12:38:00 +0000 (20:38 +0800)]
Merge pull request #9114 from pengli:dnn_rebase

add libdnn acceleration to dnn module  (#9114)

* import libdnn code

Signed-off-by: Li Peng <peng.li@intel.com>
* add convolution layer ocl acceleration

Signed-off-by: Li Peng <peng.li@intel.com>
* add pooling layer ocl acceleration

Signed-off-by: Li Peng <peng.li@intel.com>
* add softmax layer ocl acceleration

Signed-off-by: Li Peng <peng.li@intel.com>
* add lrn layer ocl acceleration

Signed-off-by: Li Peng <peng.li@intel.com>
* add innerproduct layer ocl acceleration

Signed-off-by: Li Peng <peng.li@intel.com>
* add HAVE_OPENCL macro

Signed-off-by: Li Peng <peng.li@intel.com>
* fix for convolution ocl

Signed-off-by: Li Peng <peng.li@intel.com>
* enable getUMat() for multi-dimension Mat

Signed-off-by: Li Peng <peng.li@intel.com>
* use getUMat for ocl acceleration

Signed-off-by: Li Peng <peng.li@intel.com>
* use CV_OCL_RUN macro

Signed-off-by: Li Peng <peng.li@intel.com>
* set OPENCL target when it is available

and disable fuseLayer for OCL target for the time being

Signed-off-by: Li Peng <peng.li@intel.com>
* fix innerproduct accuracy test

Signed-off-by: Li Peng <peng.li@intel.com>
* remove trailing space

Signed-off-by: Li Peng <peng.li@intel.com>
* Fixed tensorflow demo bug.

Root cause is that tensorflow has different algorithm with libdnn
to calculate convolution output dimension.

libdnn don't calculate output dimension anymore and just use one
passed in by config.

* split gemm ocl file

split it into gemm_buffer.cl and gemm_image.cl

Signed-off-by: Li Peng <peng.li@intel.com>
* Fix compile failure

Signed-off-by: Li Peng <peng.li@intel.com>
* check env flag for auto tuning

Signed-off-by: Li Peng <peng.li@intel.com>
* switch to new ocl kernels for softmax layer

Signed-off-by: Li Peng <peng.li@intel.com>
* update softmax layer

on some platform subgroup extension may not work well,
fallback to non subgroup ocl acceleration.

Signed-off-by: Li Peng <peng.li@intel.com>
* fallback to cpu path for fc layer with multi output

Signed-off-by: Li Peng <peng.li@intel.com>
* update output message

Signed-off-by: Li Peng <peng.li@intel.com>
* update fully connected layer

fallback to gemm API if libdnn return false

Signed-off-by: Li Peng <peng.li@intel.com>
* Add ReLU OCL implementation

* disable layer fusion for now

Signed-off-by: Li Peng <peng.li@intel.com>
* Add OCL implementation for concat layer

Signed-off-by: Wu Zhiwen <zhiwen.wu@intel.com>
* libdnn: update license and copyrights

Also refine libdnn coding style

Signed-off-by: Wu Zhiwen <zhiwen.wu@intel.com>
Signed-off-by: Li Peng <peng.li@intel.com>
* DNN: Don't link OpenCL library explicitly

* DNN: Make default preferableTarget to DNN_TARGET_CPU

User should set it to DNN_TARGET_OPENCL explicitly if want to
use OpenCL acceleration.

Also don't fusion when using DNN_TARGET_OPENCL

* DNN: refine coding style

* Add getOpenCLErrorString

* DNN: Use int32_t/uint32_t instread of alias

* Use namespace ocl4dnn to include libdnn things

* remove extra copyTo in softmax ocl path

Signed-off-by: Li Peng <peng.li@intel.com>
* update ReLU layer ocl path

Signed-off-by: Li Peng <peng.li@intel.com>
* Add prefer target property for layer class

It is used to indicate the target for layer forwarding,
either the default CPU target or OCL target.

Signed-off-by: Li Peng <peng.li@intel.com>
* Add cl_event based timer for cv::ocl

* Rename libdnn to ocl4dnn

Signed-off-by: Li Peng <peng.li@intel.com>
Signed-off-by: wzw <zhiwen.wu@intel.com>
* use UMat for ocl4dnn internal buffer

Remove allocateMemory which use clCreateBuffer directly

Signed-off-by: Li Peng <peng.li@intel.com>
Signed-off-by: wzw <zhiwen.wu@intel.com>
* enable buffer gemm in ocl4dnn innerproduct

Signed-off-by: Li Peng <peng.li@intel.com>
* replace int_tp globally for ocl4dnn kernels.

Signed-off-by: wzw <zhiwen.wu@intel.com>
Signed-off-by: Li Peng <peng.li@intel.com>
* create UMat for layer params

Signed-off-by: Li Peng <peng.li@intel.com>
* update sign ocl kernel

Signed-off-by: Li Peng <peng.li@intel.com>
* update image based gemm of inner product layer

Signed-off-by: Li Peng <peng.li@intel.com>
* remove buffer gemm of inner product layer

call cv::gemm API instead

Signed-off-by: Li Peng <peng.li@intel.com>
* change ocl4dnn forward parameter to UMat

Signed-off-by: Li Peng <peng.li@intel.com>
* Refine auto-tuning mechanism.

- Use OPENCV_OCL4DNN_KERNEL_CONFIG_PATH to set cache directory
  for fine-tuned kernel configuration.
  e.g. export OPENCV_OCL4DNN_KERNEL_CONFIG_PATH=/home/tmp,
  the cache directory will be /home/tmp/spatialkernels/ on Linux.

- Define environment OPENCV_OCL4DNN_ENABLE_AUTO_TUNING to enable
  auto-tuning.

- OPENCV_OPENCL_ENABLE_PROFILING is only used to enable profiling
  for OpenCL command queue. This fix basic kernel get wrong running
  time, i.e. 0ms.

- If creating cache directory failed, disable auto-tuning.

* Detect and create cache dir on windows

Signed-off-by: Li Peng <peng.li@intel.com>
* Refine gemm like convolution kernel.

Signed-off-by: Li Peng <peng.li@intel.com>
* Fix redundant swizzleWeights calling when use cached kernel config.

* Fix "out of resource" bug when auto-tuning too many kernels.

* replace cl_mem with UMat in ocl4dnnConvSpatial class

* OCL4DNN: reduce the tuning kernel candidate.

This patch could reduce 75% of the tuning candidates with less
than 2% performance impact for the final result.

Signed-off-by: Zhigang Gong <zhigang.gong@intel.com>
* replace cl_mem with umat in ocl4dnn convolution

Signed-off-by: Li Peng <peng.li@intel.com>
* remove weight_image_ of ocl4dnn inner product

Actually it is unused in the computation

Signed-off-by: Li Peng <peng.li@intel.com>
* Various fixes for ocl4dnn

1. OCL_PERFORMANCE_CHECK(ocl::Device::getDefault().isIntel())
2. Ptr<OCL4DNNInnerProduct<float> > innerProductOp
3. Code comments cleanup
4. ignore check on OCL cpu device

Signed-off-by: Li Peng <peng.li@intel.com>
* add build option for log softmax

Signed-off-by: Li Peng <peng.li@intel.com>
* remove unused ocl kernels in ocl4dnn

Signed-off-by: Li Peng <peng.li@intel.com>
* replace ocl4dnnSet with opencv setTo

Signed-off-by: Li Peng <peng.li@intel.com>
* replace ALIGN with cv::alignSize

Signed-off-by: Li Peng <peng.li@intel.com>
* check kernel build options

Signed-off-by: Li Peng <peng.li@intel.com>
* Handle program compilation fail properly.

* Use std::numeric_limits<float>::infinity() for large float number

* check ocl4dnn kernel compilation result

Signed-off-by: Li Peng <peng.li@intel.com>
* remove unused ctx_id

Signed-off-by: Li Peng <peng.li@intel.com>
* change clEnqueueNDRangeKernel to kernel.run()

Signed-off-by: Li Peng <peng.li@intel.com>
* change cl_mem to UMat in image based gemm

Signed-off-by: Li Peng <peng.li@intel.com>
* check intel subgroup support for lrn and pooling layer

Signed-off-by: Li Peng <peng.li@intel.com>
* Fix convolution bug if group is greater than 1

Signed-off-by: Li Peng <peng.li@intel.com>
* Set default layer preferableTarget to be DNN_TARGET_CPU

Signed-off-by: Li Peng <peng.li@intel.com>
* Add ocl perf test for convolution

Signed-off-by: Li Peng <peng.li@intel.com>
* Add more ocl accuracy test

Signed-off-by: Li Peng <peng.li@intel.com>
* replace cl_image with ocl::Image2D

Signed-off-by: Li Peng <peng.li@intel.com>
* Fix build failure in elementwise layer

Signed-off-by: Li Peng <peng.li@intel.com>
* use getUMat() to get blob data

Signed-off-by: Li Peng <peng.li@intel.com>
* replace cl_mem handle with ocl::KernelArg

Signed-off-by: Li Peng <peng.li@intel.com>
* dnn(build): don't use C++11, OPENCL_LIBRARIES fix

* dnn(ocl4dnn): remove unused OpenCL kernels

* dnn(ocl4dnn): extract OpenCL code into .cl files

* dnn(ocl4dnn): refine auto-tuning

Defaultly disable auto-tuning, set OPENCV_OCL4DNN_ENABLE_AUTO_TUNING
environment variable to enable it.

Use a set of pre-tuned configs as default config if auto-tuning is disabled.
These configs are tuned for Intel GPU with 48/72 EUs, and for googlenet,
AlexNet, ResNet-50

If default config is not suitable, use the first available kernel config
from the candidates. Candidate priority from high to low is gemm like kernel,
IDLF kernel, basick kernel.

* dnn(ocl4dnn): pooling doesn't use OpenCL subgroups

* dnn(ocl4dnn): fix perf test

OpenCV has default 3sec time limit for each performance test.
Warmup OpenCL backend outside of perf measurement loop.

* use ocl::KernelArg as much as possible

Signed-off-by: Li Peng <peng.li@intel.com>
* dnn(ocl4dnn): fix bias bug for gemm like kernel

* dnn(ocl4dnn): wrap cl_mem into UMat

Signed-off-by: Li Peng <peng.li@intel.com>
* dnn(ocl4dnn): Refine signature of kernel config

- Use more readable string as signture of kernel config
- Don't count device name and vendor in signature string
- Default kernel configurations are tuned for Intel GPU with
  24/48/72 EUs, and for googlenet, AlexNet, ResNet-50 net model.

* dnn(ocl4dnn): swap width/height in configuration

* dnn(ocl4dnn): enable configs for Intel OpenCL runtime only

* core: make configuration helper functions accessible from non-core modules

* dnn(ocl4dnn): update kernel auto-tuning behavior

Avoid unwanted creation of directories

* dnn(ocl4dnn): simplify kernel to workaround OpenCL compiler crash

* dnn(ocl4dnn): remove redundant code

* dnn(ocl4dnn): Add more clear message for simd size dismatch.

* dnn(ocl4dnn): add const to const argument

Signed-off-by: Li Peng <peng.li@intel.com>
* dnn(ocl4dnn): force compiler use a specific SIMD size for IDLF kernel

* dnn(ocl4dnn): drop unused tuneLocalSize()

* dnn(ocl4dnn): specify OpenCL queue for Timer and convolve() method

* dnn(ocl4dnn): sanitize file names used for cache

* dnn(perf): enable Network tests with OpenCL

* dnn(ocl4dnn/conv): drop computeGlobalSize()

* dnn(ocl4dnn/conv): drop unused fields

* dnn(ocl4dnn/conv): simplify ctor

* dnn(ocl4dnn/conv): refactor kernelConfig localSize=NULL

* dnn(ocl4dnn/conv): drop unsupported double / untested half types

* dnn(ocl4dnn/conv): drop unused variable

* dnn(ocl4dnn/conv): alignSize/divUp

* dnn(ocl4dnn/conv): use enum values

* dnn(ocl4dnn): drop unused innerproduct variable

Signed-off-by: Li Peng <peng.li@intel.com>
* dnn(ocl4dnn): add an generic function to check cl option support

* dnn(ocl4dnn): run softmax subgroup version kernel first

Signed-off-by: Li Peng <peng.li@intel.com>
7 years agocalib3d: export parameters of CALIB_CB_CLUSTERING
Pavel Rojtberg [Fri, 8 Sep 2017 14:35:23 +0000 (16:35 +0200)]
calib3d: export parameters of CALIB_CB_CLUSTERING

7 years agoMerge pull request #9752 from alalek:eigenNonSymmetric
Vadim Pisarevsky [Mon, 2 Oct 2017 11:51:13 +0000 (11:51 +0000)]
Merge pull request #9752 from alalek:eigenNonSymmetric

7 years agoMerge pull request #9756 from pranitbauva1997:doc-typo-faster
Alexander Alekhin [Mon, 2 Oct 2017 11:26:07 +0000 (14:26 +0300)]
Merge pull request #9756 from pranitbauva1997:doc-typo-faster

doc: fix typo in py_tutorials

7 years agoMerge pull request #9746 from alalek:fix_msvs2017_warning
Vadim Pisarevsky [Mon, 2 Oct 2017 11:15:48 +0000 (11:15 +0000)]
Merge pull request #9746 from alalek:fix_msvs2017_warning

7 years agoMerge pull request #9757 from berak:imgcodecs_4chan_bmp
Alexander Alekhin [Mon, 2 Oct 2017 10:27:59 +0000 (10:27 +0000)]
Merge pull request #9757 from berak:imgcodecs_4chan_bmp

7 years agodoc: fix typo in py_tutorials
Pranit Bauva [Sun, 1 Oct 2017 13:16:58 +0000 (18:46 +0530)]
doc: fix typo in py_tutorials

7 years agoUpdate HOGDescriptor documentation
Suleyman TURKMEN [Tue, 19 Sep 2017 01:28:14 +0000 (04:28 +0300)]
Update HOGDescriptor documentation

7 years agoimgcodecs: fix 4 reading channel bmp images
berak [Sun, 1 Oct 2017 13:32:55 +0000 (15:32 +0200)]
imgcodecs: fix 4 reading channel bmp images

7 years agobuild: fix MSVS2017 warning
Alexander Alekhin [Fri, 29 Sep 2017 15:16:55 +0000 (18:16 +0300)]
build: fix MSVS2017 warning

7 years agocore: cv::eigenNonSymmetric() via EigenvalueDecomposition
Alexander Alekhin [Sat, 30 Sep 2017 11:55:13 +0000 (11:55 +0000)]
core: cv::eigenNonSymmetric() via EigenvalueDecomposition

7 years agoUpdate OpenCVFindLibsGrfmt.cmake
Suleyman TURKMEN [Sat, 30 Sep 2017 18:39:50 +0000 (21:39 +0300)]
Update OpenCVFindLibsGrfmt.cmake

7 years agoMerge pull request #9743 from DensityCo:tbb-debug
Alexander Alekhin [Fri, 29 Sep 2017 14:17:55 +0000 (14:17 +0000)]
Merge pull request #9743 from DensityCo:tbb-debug

7 years agoMerge pull request #9740 from tomoaki0705:universalArithm
Alexander Alekhin [Fri, 29 Sep 2017 14:17:01 +0000 (14:17 +0000)]
Merge pull request #9740 from tomoaki0705:universalArithm

7 years agoMerge pull request #9726 from blendin:fix_oob
Alexander Alekhin [Fri, 29 Sep 2017 14:16:40 +0000 (14:16 +0000)]
Merge pull request #9726 from blendin:fix_oob

7 years agoMerge pull request #9741 from alalek:stitching_test_update
Alexander Alekhin [Fri, 29 Sep 2017 13:10:53 +0000 (13:10 +0000)]
Merge pull request #9741 from alalek:stitching_test_update

7 years agostitching(perf): update condition threshold
Alexander Alekhin [Thu, 28 Sep 2017 16:21:18 +0000 (19:21 +0300)]
stitching(perf): update condition threshold

7 years agotest: clarify fail messages
Alexander Alekhin [Thu, 28 Sep 2017 16:07:58 +0000 (19:07 +0300)]
test: clarify fail messages

7 years agouse universal intrinsic in arithm.cpp
Tomoaki Teshima [Thu, 28 Sep 2017 23:33:15 +0000 (08:33 +0900)]
use universal intrinsic in arithm.cpp
  * remove raw NEON/SSE2 implementation as much as possible
  * replace them to universal intrinsic in InRange/Compare/AddWeighted

7 years agoTBB Debug Release
bradford barr [Thu, 28 Sep 2017 21:31:41 +0000 (17:31 -0400)]
TBB Debug Release

OpenCV fails to detect tbb on a debug build if the platform has only installed
debug libraries.  This PR adds an additional check to the tbb detect logic for
systems that only install tbb debug and not both tbb debug and release.

7 years agoMerge pull request #9491 from dkurt:tf_lstm
Vadim Pisarevsky [Thu, 28 Sep 2017 21:04:06 +0000 (21:04 +0000)]
Merge pull request #9491 from dkurt:tf_lstm

7 years agoMerge pull request #9683 from terfendail:grects_weights
Vadim Pisarevsky [Thu, 28 Sep 2017 14:02:12 +0000 (14:02 +0000)]
Merge pull request #9683 from terfendail:grects_weights