Player Experience Extraction from Gameplay Video.

摘要: The ability to extract the sequence of game events for a given player's play-through has traditionally required access game's engine or source code. This serves as barrier researchers, developers, and hobbyists who might otherwise benefit from these logs. In this paper we present two approaches derive logs video via convolutional neural networks transfer learning. We evaluate in Super Mario Bros. clone, Mega Man Skyrim. Our results demonstrate our approach outperforms random forest other baselines.