Yinan Zhang

Papers

Total Citations

222

H-Index

About

Yinan Zhang is a leading researcher at the intersection of computer vision, graphics, and robotics, with a primary focus on dynamic scene representation and safe reinforcement learning. Zhang’s most impactful contribution is the introduction of **Neural Radiance Flow (NeRFlow)** in 2021, a seminal work that has garnered over 200 citations. This method extends neural implicit representations to the 4D domain, enabling the synthesis of novel views and video processing for dynamic scenes by jointly learning 3D occupancy, radiance, and temporal motion from a sparse set of RGB images. This breakthrough has significantly advanced the field of novel view synthesis for non-rigid scenes, offering a powerful tool for applications in virtual reality and video editing. In earlier work, Zhang also explored the critical challenge of safe robot learning, proposing a supervised framework that leverages a pre-existing control policy to accelerate and secure the training of new policies. By blending a supervisor’s actions with a learned policy, this approach addresses key safety concerns in reinforcement learning for physical systems. Zhang’s research is characterized by its blend of theoretical depth and practical impact, pushing the boundaries of how machines perceive and interact with dynamic, real-world environments.