Visual Semantics Events

Explore cutting-edge research in visual event classification and multilingual video analysis through this informative talk by Kate Sanders, a PhD student at Johns Hopkins University's Center for Language & Speech Processing. Dive into two groundbreaking works: the SQUID-E dataset, which addresses the challenge of ambiguous images in visual classification tasks, and MultiVENT, a multilingual video dataset for event-centric analysis. Learn how these projects aim to improve model performance on uncertain visual data and leverage diverse online news sources. Gain insights into the creation of robust datasets, the characterization of human uncertainty in vision tasks, and the development of complex multilingual video retrieval models. Discover the potential applications of these research efforts in enhancing visual event classification and multimodal information retrieval across languages.