MoE3D: A Mixture-of-Experts Module for 3D Reconstruction
By: Zichen Wang , Ang Cao , Liam J. Wang and more
Potential Business Impact:
Makes 3D pictures clearer and more real.
MoE3D is a mixture-of-experts module designed to sharpen depth boundaries and mitigate flying-point artifacts (highlighted in red) of existing feed-forward 3D reconstruction models (left side). MoE3D predicts multiple candidate depth maps and fuses them via dynamic weighting (visualized by MoE weights on the right side). When integrated with a pre-trained 3D reconstruction backbone such as VGGT, it substantially enhances reconstruction quality with minimal additional computational overhead. Best viewed digitally.
Similar Papers
PoseMoE: Mixture-of-Experts Network for Monocular 3D Human Pose Estimation
CV and Pattern Recognition
Helps computers guess a person's 3D shape from one picture.
MoRE: 3D Visual Geometry Reconstruction Meets Mixture-of-Experts
CV and Pattern Recognition
Builds 3D shapes from pictures more accurately.
3D-MoE: A Mixture-of-Experts Multi-modal LLM for 3D Vision and Pose Diffusion via Rectified Flow
Computation and Language
Helps robots understand and plan tasks in 3D.