RSD Difference of Gaussian
Research on RSD (Rotating Second Derivative) Difference of Gaussian (DOG) methods, and Gaussian Splatting more broadly, focuses on creating efficient and high-fidelity 3D scene representations, particularly for novel view synthesis and real-time rendering. Current efforts center on improving the accuracy and efficiency of Gaussian-based models, addressing challenges like overfitting, memory limitations, and handling diverse data types (e.g., LiDAR, raw images, event cameras). These advancements have significant implications for various applications, including 3D reconstruction, robotics, augmented/virtual reality, and computer vision tasks such as object pose estimation and image editing.
Papers
Sum-of-squares lower bounds for Non-Gaussian Component Analysis
Ilias Diakonikolas, Sushrut Karmalkar, Shuo Pang, Aaron Potechin
On learning higher-order cumulants in diffusion models
Gert Aarts, Diaa E. Habibi, Lingxiao Wang, Kai Zhou
Likelihood approximations via Gaussian approximate inference
Thang D. Bui
Bidirectional Recurrence for Cardiac Motion Tracking with Gaussian Process Latent Coding
Jiewen Yang, Yiqun Lin, Bin Pu, Xiaomeng Li