Yinhuai Wang

I am a third year master student at Peking University, SECE, advised by Jian Zhang. Currently, I am interning at IDEA Research, where my research focuses on physics-based human motion generation, working closely with Ailing Zeng and Lei Zhang.

Before entering Peking University, I was a start-up employee for DH-Robotics, responsible for the electronic design and motion control of the robot gripper and arm, working with Jian S Dai.

Email  /  CV  /  Google Scholar  /  Github  /  Zhihu

I'm interested in Computer Vision, Machine Learning, Robotics, and their intersections. My long-term goal is to enable robots to master all human skills and eventually build autonomous robotic systems in physical simulations and reality.

We enable physically simulated humanoids to learn interaction skills from video demonstrations, without the need of designing task-specific rewards.

We bring Range-Null space Decomposition (RND) into diffusion models, enabling diverse image restoration tasks in a zero-shot manner, without extra training or optimization.

We bring Range-Null space Decomposition (RND) into GAN-Prior based SR models to accelerate the convergence and ensure the downsampling consistency.

We fuse the generative prior and the image prior dynamically according to the degradation levels.

Extending DDNM to solve image restoration with arbitrary image size.

We build a thin-lens imaging model for NeRF, enabling it to render defocus effects.

FreeDoM is a simple but effective training-free method generating results under control from various conditions using unconditional diffusion models.

We find that simply adding an independent LN to each layer can robustly improve the performance of vision transformers.

We propose a novel framework named Null-Space Diffusion Sampling (NSDS) to solve the point cloud completion task in a zero-shot manner.

This cool template is stolen from Jon Barron!