sky area detection without deep neural networks. A c++ implemention of the algorithm mentioned in paper "Sky Region Detection in a Single Image for Autonomous Ground Robot Navigation". A fast and robust method to extract the sky area of an image.
This implementation need opencv library. Since the project is static compiled for conveniently transplanting you need static opencv library linked to this project. If static compilation is not necessary for you feel free to modify the CMakeList to dynamic compile the project.
This software has only been tested on ubuntu 16.04(x64), opencv3.4. To install this package your compiler need to support C++11.
git clone https://github.com/MaybeShewill-CV/sky-detector.git
cd ROOT_FOLDER_DIR
mkdir build
cd build
cmake ..
make -j
The project will generate a static binary file which can be used on other platform without any dynamic library. The binary file built on Ubuntu 16.04LTS was tested on CentOS 6 and worked correctly.
cd build_dir
./detect input_image_file_path output_image_file_path
Test input image with full sky
Test input image with ful sky result
Test input image with part sky
Test input image with part sky result
- Accelerate the calculation process
