When: January 25, 2024, 10h30-11h30
Where: Inria, Byron Beige (meeting room of 5th floor)
Speaker: Romain Trachel
Title: Machine learning for emotion detection in video games
Abstract: Machine learning is currently attracting a great deal of interest in video games. In this presentation, I’ll be talking about emotion detection, which I’ve applied to character animation and user research studies in the game « Marvel’s Guardians of the Galaxy » released by Eidos-Montréal in 2021. For character animation, I applied a pipeline of facial expression recognition in head-mounted video recordings of performance-capture acting. The emotions produced by the pipeline were used to create animation curves and produced more expressive conversations between the game’s main characters during gameplay sequences (i.e. non-cinematic dialogues). For user research studies, I started by analyzing facial expressions of playtest participants using a similar pipeline. Then during the pandemic, I extended the experimental setup with biometric data (electro-dermal and cardiac activity). This approach produced a set of biometric storyboards that better conveyed the emotional experience of the participants during the playtest session. The results were used to drive post-session interviews in which participants were asked to recall and describe their feelings for each emotional event detected by machine learning. To summarize, emotion detection can save resources on repetitive and time-consuming tasks in animation and it has an enormous potential to produce valuable data in playtesting. Importantly, data protection, privacy and ethics need to be regulated before deploying emotion detection in production. In addition, there’s still a long way to go before its widespread adoption as current solutions still lack confidence and robustness in the eyes of game professionals.
Short bio: Romain Trachel (PhD) is an experienced machine learning expert in the video game industry with an academic background. He has spent the last five years in AAA game studios such as Fishlabs GmbH (Hamburg, Germany), Eidos Sherbrooke and Eidos Montreal (Quebec, Canada). He obtained his PhD at the University of Nice in 2014, working on brain-computer interfaces in collaboration with Dr. Maureen Clerc in the athena team at INRIA Sophia Antipolis and Dr. Thomas Brochier at Institut de Neuroscience de la Timone (CNRS, Marseille). Over the course of his career, he has acquired solid expertise in machine learning, signal and image processing and cognitive neuroscience, as well as professional experience in the video game industry.
