The paper proposes DDN-SLAM, a real-time dense neural implicit semantic SLAM system designed for dynamic scenes. It addresses the challenges faced by existing neural implicit SLAM systems in real-world environments with dynamic interferences. DDN-SLAM utilizes the priors provided by the deep semantic system and conditional probability fields for segmentation. It ensures fast hole filling and high-quality mapping while mitigating the effects of dynamic information interference. DDN-SLAM supports monocular, stereo, and RGB-D inputs, operating robustly at a frequency of 20-30Hz. The method outperforms state-of-the-art approaches in both dynamic and static scenes.
Publication date: 5 Jan 2024
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2401.01545
