Introduction to OpenCL of opencv

Intro

Introduce

Open Computing Language (OpenCL) is an open standard for writing code that runs across heterogeneous platforms including CPUs, GPUs, DSPs and etc. In particular OpenCL provides applications with an access to GPUs for non-graphical computing (GPGPU) that in some cases results in significant speed-up. In Computer Vision many algorithms can run on a GPU much more effectively than on a CPU: e.g. image processing, matrix arithmetic, computational photography, object detection etc.

Open computing language （OpenCL） It's an open standard , Used to write cross heterogeneous platforms （ Include CPU、GPU、DSP etc. ） Running code . In especial ,OpenCL Provides support for applications GPU The interview of , For non graphical calculations （GPGPU）, In some cases, it will significantly increase the speed . In computer vision , Many algorithms can be used in GPU Compared with CPU Run more effectively on ： For example, image processing 、 Matrix operations 、 Computational photography 、 Target detection, etc .

History

history

Acceleration of OpenCV with OpenCL started 2011 by AMD. As the result the OpenCV-2.4.3 release included the new oclmodule containing OpenCL implementations of some existing OpenCV algorithms. That is, when OpenCL runtime and a compatible device are available on a client machine, user may call cv::ocl::resize() instead of cv::resize() to use the accelerated code. During 3 years more and more functions and classes have been added to the ocl module; but it has been a separate API alongside with the primary CPU-oriented API in OpenCV-2.x.

​AMD from 2011 Year started OpenCL Speed up OpenCV. therefore ,OpenCV-2.4.3 The version includes the new OCL modular , It includes some existing OpenCV Algorithm OpenCL Realization . in other words , When there is... On the client computer OpenCL Runtime and compatible devices , Users can call cv::ocl::resize（） instead of cv::resize（） To use accelerated code . For three years ,ocl More and more functions and classes have been added to the module ; But it has something to do with OpenCV-2 The main aspects of CPU Of API Related independence .​

In OpenCV-3.x the architecture concept has been changed to the so-called Transparent API (T-API). In the new architecture a separate OpenCL-accelerated cv::ocl::resize() is removed from external API and becomes a branch in a regular cv::resize(). This branch is called automatically when it’s possible and makes sense from the performance point of view. The T-API implementation was sponsored by AMD and Intel companies.

​ stay OpenCV-3.x in , The concept of architecture has been changed to so-called transparency API（T-API）. In the new architecture , A single OpenCL Speed up cv::ocl::resize（） From the outside API Delete in , And become routine cv::resize（） A branch of . Where possible , This branch will be called automatically , And from a performance point of view , This branch is meaningful .T-API The realization of is by AMD And Intel sponsored .

Numbers

Some performance numbers are shown on the picture below:

Some performance data are shown in the figure below ：

Code sample

Code example

Regular CPU code

routine CPU Code

// initialization
VideoCapture vcap(...);
CascadeClassifier fd("haar_ff.xml");
Mat frame, frameGray;
vector<rect> faces;
for(;;){
  // processing loop
  vcap >> frame;
  cvtColor(frame, frameGray, BGR2GRAY);
  equalizeHist(frameGray, frameGray);
  fd.detectMultiScale(frameGray, faces, ...);
  // draw rectangles …
  // show image …
}

OpenCL-aware code OpenCV-2.x

Support OpenCL Code for OpenCV-2.x

// initialization
VideoCapture vcap(...);
ocl::OclCascadeClassifier fd("haar_ff.xml");
ocl::oclMat frame, frameGray;
Mat frameCpu;
vector<rect> faces;
for(;;){
  // processing loop
  vcap >> frameCpu;
  frame = frameCpu;
  ocl::cvtColor(frame, frameGray, BGR2GRAY);
  ocl::equalizeHist(frameGray, frameGray);
  fd.detectMultiScale(frameGray, faces, ...);
  // draw rectangles …
  // show image …
}

OpenCL-aware code OpenCV-3.x

Support OpenCL Code for OpenCV-3.x

// initialization
VideoCapture vcap(...);
CascadeClassifier fd("haar_ff.xml");
UMat frame, frameGray;
vector<rect> faces;
for(;;){
  // processing loop
  vcap >> frame;
  cvtColor(frame, frameGray, BGR2GRAY);
  equalizeHist(frameGray, frameGray);
  fd.detectMultiScale(frameGray, faces, ...);
  // draw rectangles …
  // show image …
}