Saddha Santanaporn
Software Development
«
Using Computer Vision to Capture Data from Static UI Elements in
League of Legends
The goal of my thesis was to research proprietary access or data streams
collect, and analyze gameplay data This rich dataset can be used for many
and implement a methodology to extract,
from League of Legends videos via an
automated computer vision and offline
post-processing techniques. Since there
were many data points to extract from
LoL, two different thesis topics were
purposes, such as helping professional
teams remain competitive by allowing
them to study their own and their
opponents’ past performance.
chosen. I used template matching and I chose this project because I wanted to
from static User Interface elements. To to game development. It took around four
feature matching to extract information
track players’ positions on a dynamic
minimap, another student (Kevin Nappoly)
utilized object detection via Convolutional
Neural Network (CNN).
The combination of two thesis projects
presented an automated offline tool
that extracts player positions, player
health and mana over time. Game time
and champions are picked through
offline techniques and don’t rely on any
94
between the LoL client and server.
SOFTWARE DEVELOPMENT
experiment in a field not directly related
months to finish an acceptable artifact
and a paper ready to be published.