Enhancing SplaTAM Performance Through Dynamic Learning Rate Decay and Optimized Keyframe Selection

• Published in: IEEE International Conference on Robotics and Biomimetics (Online) pp. 1972 - 1977
• Main Authors: Ma, Xingwang, Wu, Xinzhao, Zhang, Liwei, Shen, Shunxi
• Format: Conference Proceeding
• Language: English
• Published: IEEE 10.12.2024
• Subjects: Cameras; Optimization; Rendering (computer graphics); Robot vision systems; Simultaneous localization and mapping; Termination of employment; Three-dimensional displays; Training; Trajectory tracking; Visualization

Dense Simultaneous Localization and Mapping (SLAM) is crucial for robotics and augmented reality applications. In this paper, we optimized the SplaTAM using 3D Gaussian representations to achieve high-quality reconstruction using RGB-D cameras. We employ an online tracking and mapping system specifi...