Neural 3D Representation
Neural 3D representation aims to create digital 3D models from 2D images or other data, enabling photorealistic rendering and manipulation. Current research focuses on improving efficiency and control, employing architectures like Gaussian Splatting and refining NeRFs (Neural Radiance Fields) through techniques such as diffusion models and incorporating dense depth information for more accurate reconstructions. These advancements are driving progress in robotics, virtual and augmented reality, and other fields requiring accurate and manipulable 3D scene understanding, particularly in scenarios with limited viewpoints or single-image input.
Papers
October 17, 2024
September 4, 2024
May 28, 2024
March 27, 2024
December 9, 2023
September 11, 2023
April 27, 2023
April 25, 2023
November 21, 2022
April 7, 2022