genetic-images/README.org
Phuntsok Drak-pa 4a5a131ef0 README update
2019-03-19 17:32:41 +01:00

83 lines
3.6 KiB
Org Mode

[[http://spacemacs.org][file:https://cdn.rawgit.com/syl20bnr/spacemacs/442d025779da2f62fc86c2082703697714db6514/assets/spacemacs-badge.svg]]
* Genetic image generation
This project is a university assignment that aims at regenerating a reference
image from random shapes of random color applied. There will be lots of
different tests on what method is the best and/or the fastest to get a new
image as close as possible to the reference image.
The first method will be the random addition of squares of random size to random
locations of random color. If the picture with the new random square improves in
similarity with the reference image, then it is preserved, otherwise the change
is discarded. This method is here as a reference for the other methods.
The second method will be the same as the first method, however The generation
of the new pictures will be multi-threaded; if two or more threads return an
improved picture, the best one will be selected. Then it will be reused by the
next threads.
The third method will instead delegate different parts of the picture to
different threads. For instance, if the software runs on four threads, the
reference picture will be divided in four zones, each one will be generated by a
separate thread.
The fourth method, which can be combined with the first three, will limit the
choice in terms of random color to the colors already existing in the reference
image.
The fifth method, which can be combined with the four first methods, will force
the random squares to be larger during the first iterations and progressively
smaller until the smallest specified size which should be reached by the last
iteration.
More methods will be added later once these four will be implemented.
* How to build this project
To build the project, you will also be required to install the development
libraries your distro offers for GTK+3. For Ubuntu, it is ~libgtk3-devel~, for
Void Linux it is ~gtk3-devel~ and for Arch Linux it is already bundled with the
~gtk3~ package.
You will also ned to have Ninja and Conan installed. To install Ninja, install
the appropriate package offered by your package manager (~ninja-build~ on
Debian, ~ninja~ on Arch Linux and Void Linux), and to install conan, use ~pip~.
#+begin_src shell
pip install --user conan
#+end_src
This project is built with conan, ninja and cmake using clang-7 for C++17. To
use it, first install clang-7 and lldb 7, then run this:
#+begin_src shell
conan profile new default --detect
conan profile update settings.compiler=clang default
conan profile update settings.compiler.version=7.0 default
conan profile update settings.compiler.libcxx=libstdc++11 default
conan profile update env.CC=/bin/clang default
conan profile update env.CXX=/bin/clang++ default
#+end_src
If you do not wish to overwrite your ~default~ profile, you can instead create a
new one, for instance ~clang~. To do so, write the name of your new profile (in
this example ~clang~) instead of ~default~ in the commands shown above.
Then, To build and run the program, go to the root of the project and run this:
#+begin_src shell
mkdir build && cd build
conan install .. --build missing
cmake -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ .. -G Ninja
cmake --build .
#+end_src
If you want to use another profile than your default one, you should run the
following line instead of the second line:
#+begin_src shell
conan install .. --build missing --profile <your_profile>
#+end_src
This project was built and tested using clang-7, lldb and gdb on Void Linux
(kernel 4.19) and Arch Linux (kernel 5.0).
* Credits
Awesome C++ Template by [[https://github.com/devkoriel/AwesomeCppTemplate][devkoriel]].