The article presents NGEL-SLAM, a system that uses neural implicit representations to provide dense geometry in Simultaneous Localization and Mapping (SLAM). Traditional SLAM systems can provide low latency and high precision tracking, but lack dense geometry and texture information. Recent advances in neural implicit representations have enabled accurate and dense 3D surface reconstruction. NGEL-SLAM combines the tracking accuracy of traditional SLAM systems with the capability of neural implicit representations to extract dense meshes and generate high-fidelity images. The system achieves state-of-the-art tracking and mapping accuracy while maintaining low latency.
Publication date: 20 Nov 2023
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2311.09525
