- Download a dataset:
- UC Merced Land Use Dataset: http://weegee.vision.ucmerced.edu/datasets/landuse.html
- Intel Image Classification: https://www.kaggle.com/puneet6060/intel-image-classification
- The Oxford-IIIT Pet Dataset: https://www.robots.ox.ac.uk/~vgg/data/pets/
- Install dependencies -- install_dependencies.sh (Minimum required OpenCV version is 4.51. You can install OpenCV 4.51 using the information at http://milq.github.io/install-opencv-ubuntu-debian/)
- Go to the DPDA code directory
- You must create two folders named "images" and "results" in "build" directory (They are provided)
- Create a folder under "images" named "train"
- Create as many folders for your classes as you wish under the "train" folder (Sample data is provided).
- Copy images to corresponding class folders -- png, jpg, and tiff are supported
- Build and run using the following commands:
- Go to "build" folder: cd build
- cmake ..
- If everything ok:
make -j 4
or
make -j 8
or
make -j 16
(change the number after j with the number of cpu cores for your system: 2-4-6-8-10-12-16-20) - Make the "DPPDA_Executable" as executable file: sudo chmod +x DPPDA_Executable
- Run executable: ./DPDA_Executable
- You will see the resulting augmented images under the results folder.
- Augmentation settings can be adjusted by using the informative comments in the "main.cpp" file.