2019/05/31 by Charles Ringer, Ringer, Charles, James Alfred Walker +3 · 1 citation
Computer Science · Mathematics · #Computational Physics and Python Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Tensor decomposition and applications
paper · pdf · doi:10.48550/arxiv.1905.13694
openalex publication_date 2019/05/31 · openalex created_date 2022/07/23 · openalex updated_date 2026/07/28
Video game streaming provides the viewer with a rich set of audio-visual\ndata, conveying information both with regards to the game itself, through game\nfootage and audio, as well as the streamer's emotional state and behaviour via\nwebcam footage and audio. Analysing player behaviour and discovering\ncorrelations with game context is crucial for modelling and understanding\nimportant aspects of livestreams, but comes with a significant set of\nchallenges - such as fusing multimodal data captured by different sensors in\nuncontrolled ('in-the-wild') conditions. Firstly, we present, to our knowledge,\nthe first data set of League of Legends livestreams, annotated for both\nstreamer affect and game context. Secondly, we propose a method that exploits\ntensor decompositions for high-order fusion of multimodal representations. The\nproposed method is evaluated on the problem of jointly predicting game context\nand player affect, compared with a set of baseline fusion approaches such as\nlate and early fusion.\n