Computer Vision News 28 Congrats, Doctor Max! Max Zhaoshuo Li recently completed his PhD at Johns Hopkins University (JHU), working with Russell H. Taylor and Mathias Unberath. His research focused on developing a vision-guided system for surgery, which reconstructs the 3D structures densely and tracks the moving objects accurately. He also worked on developing infrastructures for next-generation mixed reality and autonomous systems. He will be joining NVIDIA as a research scientist to contribute to the development of 3D AI systems. Congrats, Doctor Max! Metaverse? Mixed reality? Autonomous systems? All these applications require a system that can understand the 3D motion and structures of a scene. The computer vision community has a strong interest in video-based solutions due to the widespread availability of cameras. Such a solution could revolutionize healthcare systems by aiding clinicians and surgeons with decision making. However, accurately and reliably recovering 3D information still poses a challenge. In his PhD dissertation, Max addresses these challenges with innovative computer vision algorithms. His algorithm named Stereo Transformer (Li et al, ICCV 2021) is the first to use attention mechanisms to achieve robust out-out-distribution depth estimation, even when trained only on synthetic dataset. Stereo Transformer is well-suited for scenarios where collecting data is difficult, such as in surgical scenes. Furthermore, the reasoning process of the Stereo Transformer can be examined visually, making it a white boxed algorithm.
RkJQdWJsaXNoZXIy NTc3NzU=