A Genetic Algorithm (GA) example based on using correlation of a selected image region against all other regions in the image to find the location of selection. Implemented as a C++ header class (ga.h) with an interface (ga_interface.cpp) to allow for variable population size (P), mutation (M) and cross-over (C) probability settings.
Illustrates all of the key elements of a GA approach including:
- genes / phenotypes: (xi,yi) locations of image regions (i.e. chromosomes / genetic representation)
- population: random set of (X x Y) dimension regions from the image each at (xi,yi).
- fitness function: correlation between mouse selected (click and drag) image region and each region in the population.
- selection: fitness proportional selection, roulette wheel.
- cross-over: uniform cross-over based on crossing xi and yi co-ordinates between two selected regions.
- mutation: randomly perturb bit-pattern of xi and yi co-ordinates of each region in the population.
-- within a readily adaptable C++ class structure.
Requires OpenCV - all tested with OpenCV 2.4.x / 3.x and GCC (Linux) and known to work with MS Visual Studio 20xx on Win32 / Win64.
Example video (click to watch):
If I taught you between 2010 and 2013 at Cranfield University or ESTIA - this is the C++ GA example from class.
Demo source code is provided "as is" to aid your learning and understanding.
Download each file as needed or to download the entire repository and run each try:
git clone https://github.com/tobybreckon/correlation-ga.git
cd correlation-ga
cmake .
make
./correlation_ga
< use sliders to select population size (P), mutation (M) and cross-over (C) probability settings >
< use mouse to click/drag to select image region >
< hold down - x key to exit / r key to reset >
The GA example runs with a webcam connected or from a command line supplied video file of a format OpenCV supports on your system (otherwise edit the code to provide your own image source).
N.B. you may need to change the line near the top that specifies the camera device to use on this example - change "0" if you have one webcam, I have to set to "1" to skip my built-in laptop webcam and use the connected USB camera.
If referencing this example in your own work please use:
@TechReport{breckon2010,
author = {Breckon, T.P. and Barnes, S.E.},
title = {Machine Learning - MSc Course Notes},
institution = {Cranfield University},
year = {2010},
address = {Bedfordshire, UK},
}
If you find any bugs please raise an issue (or better still submit a pull request, please) - toby.breckon@durham.ac.uk
"may the source be with you" - anon.