The aim of this project was to write an application to perform Pepsi drinks logos detection. It was written fully in C++17 with a small help of OpenCV (only cv::imread, cv::imwrite and cv::Mat* are used). All of the image processing algorithm was written manually. Application works rather cross-platform, in command line, and provides a small subset of unit tests. Images used in logo detection were taken both from internet and manually, using my smartphone's camera. Algorithms parameters may be tuned using dedicated JSON files.
Except of course an OpenCV library, in order to finish the project, I've decided to use following cool stuffs:
CMake- to manage build process,Conan- to handle 3rd party dependencies,spdlog- excellent library for logging,Catch2- unit testing in BDD,jsonformoderncpp- reading/writing JSON files,CLI11- setting up a CLI interface.
- Clone the repository,
- Create and go to build directory:
mkdir build/ && cd build, - Install conan dependencies:
conan install .., - Configure CMake:
cmake .. -DCMAKE_BUILD_TYPE=<build_type>, - Build:
make -j4, - Run the program and pass image to it:
./bin/find_logos assets/camera/<n>.jpg, where<n>is number of image, - If you would like to see all steps of image processing, provide also
-v(--verbose) option, - Optionally - run unit tests:
./bin/detector_test.