visioncpp

Fast, GPU-accelerated computer vision and image processing


Keywords
vision, image, processing, gpu, sycl, computecpp, machine, learning
License
Apache-2.0
Install
pip install visioncpp==0.1.0

Documentation



Overview

VisionCpp is a lightweight header-only library for computer vision and image processing. The aim of the library is to provide a toolbox that enables performance portability for heterogeneous platforms using modern C++.

Written using SYCL 1.2 and compiled/tested with ComputeCpp to accelerate vision code using OpenCL devices.

Table of contents

Integration

All you need to do is include the VisionCpp.hpp header in your project and you are good to go! ( assuming that OpenCL and ComputeCPP is installed correctly. )

#include <visioncpp.hpp> //all that is needed

Sample Code

Below is a very simple application that will do the conversion RGB -> HSV. Full source code can be found in the examples folder. RGB is assumed to be a three-channel unsigned char storage with a reasonable channel order.

  // main, args, checks and all the boring stuff

  // ...

  // where VisionCpp will run.
  auto dev = visioncpp::make_device<visioncpp::backend::sycl,
                                    visioncpp::device::cpu>();

  // create a host container for input data
  std::shared_ptr<unsigned char> in_rgb(new unsigned char[3],
  [](unsigned char *dataMem) { delete[] dataMem;});

  in_rgb.get()[0] = atoi(argv[1]);
  in_rgb.get()[1] = atoi(argv[2]);
  in_rgb.get()[2] = atoi(argv[3]);

  // create a host container for output data
  std::shared_ptr<unsigned char> out_hsv(new unsigned char[3],
  [](unsigned char *dataMem) { delete[] dataMem;});

  // exiting this scope will sync data
  {
    // definition of the VisionCpp pipeline:

    // create terminal nodes - a leaf node ( data node ) of the expression tree.
    // terminal struct takes 4 arguments
    // 1st template parameter specifies the data U8 (unsigned char) C3 (three
    // channels)
    // 2nd number of columns in the storage
    // 3rd number of rows in the storage
    // 4th underlying storage type - currently only Buffer2D supported
    auto data =
        visioncpp::terminal<visioncpp::pixel::U8C3, 1, 1,
                            visioncpp::memory_type::Buffer2D>(in_rgb.get());
    auto data_out =
        visioncpp::terminal<visioncpp::pixel::U8C3, 1, 1,
                            visioncpp::memory_type::Buffer2D>(out_hsv.get());

    // unsigned char -> float RGB storage conversion
    auto node = visioncpp::point_operation<visioncpp::OP_U8C3ToF32C3>(data);
    // float RGB to float HSV conversion
    auto node2 = visioncpp::point_operation<visioncpp::OP_RGBToHSV>(node);
    // helper node that allows display of HSV
    // for unsigned char: V <- 255*V, S <- 255*S, H <- H/2 ( to fit in range of 0..255 )
    auto node3 = visioncpp::point_operation<visioncpp::OP_HSVToU8C3>(node2);

    // assign operation that writes output of the pipe to output terminal node
    auto pipe = visioncpp::assign(data_out, node3);
    // execute the pipeline
    // 1st template parameter defines if VisionCpp back-end fuses the expression
    // 2nd & 3rd shared memory sizes ( column, row )
    // 4th & 5th local work group size ( column , row )
    visioncpp::execute<visioncpp::policy::Fuse, 1, 1, 1, 1>(pipe, dev);
  }

  printf("RGB: %u %u %u \nHSV: %u %u %u \n", in_rgb.get()[0], in_rgb.get()[1],
         in_rgb.get()[2], out_hsv.get()[0], out_hsv.get()[1], out_hsv.get()[2]);

Requirements

To successfully compile VisionCpp tests, you will need:

Build

Assuming you are in the root of a git repo:

mkdir build
cd  build
cmake .. -DCOMPUTECPP_PACKAGE_ROOT_DIR={PATH_TO_COMPUTECPP_ROOT} -DCMAKE_CXX_COMPILER={FAVORITE_CXX_COMPILER}
make -j8
make test

The output binaries will be catalogued in bin folder.

| - build
  | - bin
    | - example
    | - test

Examples

Most of the examples are using camera.

Documentation

Online documentation can be found here.

The documentation is created using Doxygen.

make doc

The documentation will be created in html folder in build directory.

| - build
  | - doc

Contributing

Contributors always welcome! See CONTRIBUTING.md for details.

The list of contributors.

Resources

License

The Apache License, Version 2.0 License. See LICENSE for more.

Known Issues

  • The Tuple class works only with clang++.