Title: Foundation Vision Models and Applications

Abstract:Foundation models become emerging research topics in the field of artificial intelligence nowadays, which have advanced the state-of-the-arts in many computer vision and natural language processing tasks. Foundation models are also key techniques for many important applications such as visual surveillance, autonomous driving, and intelligent devices. This talk will review the recent research progress of foundation vision models from the perspectives of model architecture and learning paradigm, and introduce some work conducted by the Intelligent Vision Group at Tsinghua University, including dynamic sparse models, global filtering models, spherical fractal models, and geometry-aware models, and their applications in various vision tasks such as object detection and segmentation, image and video retrieval, and 3D reconstruction and recognition.