ICCV Daily 2021 - Friday

“ Microsoft Research give us freedom. We’re not limited to specific tasks – we can work on anything we’re interested in, ” he reveals. “ As we’re given so much freedom, we want to identify what the most important things to explore are and really push those. For now, we’re mainly focused on two things. One is how to create a proper architecture for computer vision which is scalable – this is what this paper is about. The second is about the future way to do learning. There have been several promising ways, including self-supervised learning, classification-based supervised learning, and multi-modality learning, but their relationship and what the best practice for learning is unknown. We want to study this direction. ” If Swin Transformer could help you in your own vision tasks, you can find the source code and pre-trained models on Github. To learn more about Han’s work [Paper ID 7354], you can view his pre- recorded oral presentation and PDF poster now and join his live Q&A session today [8B] at 16:00-18:00 EDT. 7 DAILY ICCV Friday “There have been several promising ways, including self- supervised learning, classification-based supervised learning, and multi-modality learning, but their relationship and what the best practice for learning is unknown. We want to study this direction.” Han Hu