News

VIULab releases TCG-AR, an open-source system that augments live trading card game broadcasts using only commodity hardware. From standard webcam feeds, the system detects every card on the play mat, identifies it against a reference database of 20,360 Pokémon TCG cards, and composites the corresponding animated sprite onto the card across multiple camera views in real time, at 30 fps on a single consumer GPU. Card identification relies on a retrieval-based pipeline that generalizes to newly released cards without retraining, reaching 96% accuracy in deck-constrained settings. The system integrates directly with OBS and includes a full broadcast overlay for commentated streams. The code, trained models, and evaluation dataset are publicly available on GitHub, and the accompanying paper is on arXiv. This work was developed by Anthony Cioppa, Antoine Verdonck, Maxim Henry, Marc Van Droogenbroeck, and Raphaël La Rocca.