Data collection with screen capture

Jason Traish, James Tulip, Wayne Moore

Research output: Book chapter/Published conference paperConference paperpeer-review

Abstract

Game traces are an important aspect of analysing how players interact with computer games and developing case based reasoning agents for such games. We present a computer vision based approach using screen capture for extracting such game traces. The system uses image templates of to identify and log changes in game state. The advantage of the system is that it only captures events which actually occur in a game and is robust in the face of multiple redundant commands and command cancellation. This paper demonstrates the use of such a vision based system to gather build orders from Starcraft 2 and compares the results generated with those produced by a system based on analysing log files of user actions. Our results show that the vision based system is capable of not only automatically retrieving data via screen capture, but does so more accurately and reliably than a system relying completely on recorded user interactions. Screen capture also allows access to data not otherwise available from an application. We show how screen capture can be used to retrieve data from the DotA 2 picking phase in real time. This data can be used to support meta-game activity, and guide in-game player behaviours.
Original languageEnglish
Title of host publicationAISB Convention 2015
Place of PublicationUnited Kingdom
PublisherThe Society for the Study of Artificial Intelligence and the Simulation of Behaviour
Pages1-4
Number of pages4
Publication statusPublished - 2015
EventArtifical Intelligence and Simulation of Behaviour Convention - University of Kent, Canterbury; United Kingdom, United Kingdom
Duration: 20 Apr 201522 Apr 2015

Conference

ConferenceArtifical Intelligence and Simulation of Behaviour Convention
CountryUnited Kingdom
Period20/04/1522/04/15

Fingerprint Dive into the research topics of 'Data collection with screen capture'. Together they form a unique fingerprint.

Cite this