This program is intended to create image training datasets for machine and deep learnings applications.
Given an input and output directory this code generates N random transformations for each image of the input directory. The patch size used is fixed to the same as the images at input directory.
This code requires OpenCV!
In this case we want to train a convolutional neural network (CNN) for keypoint matching task. Given the patches at input dir this code generate the following "random" transformations for them.
mkdir build
cd build
cmake ..
make
./create_dataset path_input path_out N