Kevin Nappoly
Software Development
«
Gameplay Data Extraction for League of Legends Through
Computer Vision
Game analytics plays an important role in learn the features within images to make
and electronic sports (esports) by getting esports much easier. This system adds
guiding teams in both traditional sports
a better understanding of the opponent’s
strategies. Unfortunately, not every esport
game has a reliable method to extract,
various visuals to games watched at home
like football or basketball.
collect and analyze the data needed for This thesis brings these two worlds
League of Legends, a popular multiplayer extract, collect, and analyze gameplay
team and player analysis. For instance,
online battle arena video game, does not
provide the ability to track gameplay data
like players’ position, hit points, and mana
within the game without having to watch
and manually encode the entire match.
On the other hand, machine learning is
taking the world stage by using object
detection to help Artificial Intelligence (AI)
86
analysis of various traditional games and
SOFTWARE DEVELOPMENT
together and presents a methodology to
data from League of Legends match
videos via an automated computer vision
system. I tracked the positions of players
on the minimap using object detection via
Convolutional Neural Networks (CNNs)
and encoded them into a data file that can
accelerate the work of analysts and will
thus lead to more informed and decisive
strategies.