MeGS-SLAM: Memory Efficient Gaussian Splatting SLAM with Graph Signal Processing

Sude Zhang1*, Zhiyong Zhang1*,
1Sun Yat-sen University
Paper Coming Soon

Abstract

Experiments performed on synthetic and real-world datasets show that our method achieves over 2× compression in memory usage and increases nearly 250% rendering speed while maintaining tracking and mapping performance.

Ablation Study on ScanNet Dataset

                                                  Ablation on edge priors                                                                                                                      Ablation on GSP

w/ edge priors
w/o edge priors
w/ GSP
w/o GSP
w/ GSP
w/o GSP

Exprimental Results

                                                                                           Replica Dataset

It can be observed that the estimated camera poses (blue frustums and trajectories) accurately align with the ground truth camera poses (red frustums and trajectories).

                                                                           Room0                                                                                                                                       Office0

Ours
SplaTAM
Ours
SplaTAM

             

           TUM-RGBD Dataset

             

                                                                         ScanNet Dataset