Based primarily in OpenCV by feeding images or video
Detects circles to run champion recognition on them. Since champion icons can be on top of each other to the point of making some of them unrecognizable even for humans. Thus, we need to work with all the knowledge that we can.
There are 3 norms that dictate which champ icons are on top on the others. From how on we are going to refer to this aspect of the code/game Icon Priority:
Using this we first try to recognize the support if an icon is on top of another and so on. The goal is keep reducing the search space with the information that is gathered. The most simple example of this is trying to recognize the blue champs only inside of blue circles and the red champs with the red circles.
make
to build the executable. By default is called opencv_lol
and is located in the project directory.Build:
nix build github:tomiock/LeagueOfLegends-Analytics
and install for the local user:
nix profile install .
Execute this commands on the project directory.
The repo can also be cloned. Use nix flake show
on the repo directory to see the available outputs. There is a shell available in which make
can be executed to build directly without using nix. Doing this will generate the compile_commands.json
correctly, unlike nix build
.
Steps:
About 80%-90% accuracy.
Deprecated for a more simple solution. I do not know what to do with this.
python3 reference_images/request_champion_square_images.py
python3 create_dataset.py
The dataset
folder can be created easily with mkdir dataset
.