ICCV Daily 2021 - Wednesday

Finally, she adds: “ I’m also raising a toddler – I don’t know if that counts but it’s sort of fun to combine that with the pandemic and working from home and everything else! ” You are all warmly invited to attend the 4th Workshop on Closing the Loop Between Vision and Language, which will be held all day on Sunday at ICCV. 12 DAILY ICCV Wednesday Workshop Challenges The Video-and-Language Understanding Evaluation (VALUE) Competition is a new multitask competition for video and language understanding, consisting of 8 datasets and 11 tasks. These tasks fall into three major categories: Question Answering, Retrieval, and Captioning. Video-and-language understanding is challenging, as it involves visual and language semantic understanding, spatio- temporal grounding, multi-modal fusion, and commonsense reasoning. The Condensed Movies Challenge uses the Condensed Movies Dataset. This contains key scenes (2-3 minutes each) from over 3000 movies with a corresponding high level semantic description for each, detailing characters, motivations, interactions and relationships. The dataset is ideal for testing long- range understanding of high-level narrative structures in movies.