potok/opencv
1
0
Fork 0
mirror of https://github.com/opencv/opencv.git synced 2026-07-31 00:03:03 +04:00
Code Packages Releases Wiki Activity

Merge pull request #9330 from hrnr:akaze_ocl

Browse Source 
[GSOC] Enable OCL for AKAZE (#9330)

* revert e0489cb - reenable OCL for AKAZE

* deal with conversion internally in AKAZE

* pass InputArray directly to AKAZE to allow distiguishing input Mat/UMat. deal with conversion there
* ensure that keypoints orientations are always computed. prevents misuse of internal AKAZE class.

* covert internal AKAZE functions to use InputArray/OutputArray

* make internal functions private in AKAZE

* split OCL and CPU paths in AKAZE

* create 2 separate pyramids, 1 for OCL and 1 for CPU
* template functions that use temporaries to always store them as correct type (UMat/Mat)

* remove variable used only in OCL path

causes unused variable warning

* update AKAZE documentation

* run ocl version only when ocl is enabled

* add tests for OCL path in AKAZE

* relax condition for keypoints angle
This commit is contained in:
Jiri Horner
2017-08-16 18:46:11 +02:00
committed by Alexander Alekhin
parent 3a8dbebd37
commit a5b5684670
5 changed files with 341 additions and 175 deletions
Download Patch File Download Diff File Expand all files Collapse all files
modules/features2d/include/opencv2/features2d.hpp
+15 -6
View File
@@ -658,13 +658,22 @@ public:
CV_WRAP virtual int getDiffusivity() const = 0;
};
/** @brief Class implementing the AKAZE keypoint detector and descriptor extractor, described in @cite ANB13 . :
/** @brief Class implementing the AKAZE keypoint detector and descriptor extractor, described in @cite ANB13.
@note AKAZE descriptors can only be used with KAZE or AKAZE keypoints. Try to avoid using *extract*
and *detect* instead of *operator()* due to performance reasons. .. [ANB13] Fast Explicit Diffusion
for Accelerated Features in Nonlinear Scale Spaces. Pablo F. Alcantarilla, Jesús Nuevo and Adrien
Bartoli. In British Machine Vision Conference (BMVC), Bristol, UK, September 2013.
*/
@details AKAZE descriptors can only be used with KAZE or AKAZE keypoints. This class is thread-safe.
@note When you need descriptors use Feature2D::detectAndCompute, which
provides better performance. When using Feature2D::detect followed by
Feature2D::compute scale space pyramid is computed twice.
@note AKAZE implements T-API. When image is passed as UMat some parts of the algorithm
will use OpenCL.
@note [ANB13] Fast Explicit Diffusion for Accelerated Features in Nonlinear
Scale Spaces. Pablo F. Alcantarilla, Jesús Nuevo and Adrien Bartoli. In
British Machine Vision Conference (BMVC), Bristol, UK, September 2013.
*/
class CV_EXPORTS_W AKAZE : public Feature2D
{
public: