Pocket-SLAM: Rendering-Area-Aware Pruning for Memory-Efficient 3DGS-SLAM
Leshu Li, Jie Peng, Yang Zhao
AI summary
Problem
3DGS-SLAM systems accumulate millions of Gaussian points in large-scale scenes, causing prohibitively high peak runtime memory consumption that blocks deployment on edge devices. Existing pruning methods rely on local heuristics that fail to preserve critical rendering information or address runtime memory bottlenecks.
Approach
The method prunes Gaussians based on their effective pixel coverage in the rendered image, guided by a tile-level budget mechanism that adaptively allocates survival budgets to prevent over-pruning in texture-dense or sparse regions.
Key results
- Over 60% reduction in peak runtime memory consumption
- More than 2× improvement in frames per second
- Preserves camera tracking and rendering accuracy comparable to unpruned baselines
- Demonstrated on large-scale outdoor KITTI and EuRoC datasets
Why it matters
Enables real-time, high-fidelity 3DGS-SLAM deployment on memory-constrained edge devices for autonomous driving and drone navigation.
Abstract
3D Gaussian Splatting (3DGS) has garnered sig- nificant attention in Simultaneous Localization and Mapping (SLAM) due to its advances in capturing fine-grained geometry features and synthesizing novel views. For SLAM in large- scale scenes, such as autonomous driving, 3DGS-SLAM faces a critical limitation. The memory consumption increases con- tinuously over time as Gaussian points accumulate, leading to poor memory efficiency and limiting its applicability. In this work, we propose a rendering-area–aware pruning strategy that selectively removes Gaussians based on their contribution to the effective rendering area, rather than solely relying on Gaussian- level heuristics (e.g., opacity or gradient magnitude). This perspective directly targets the sources of memory redundancy, effectively reducing the peak memory footprint of 3DGS- SLAM during runtime. Evaluations on the EuRoC and KITTI datasets demonstrate that our method consistently outperforms existing pruning approaches in large-scale outdoor scenes, achieving over 60% memory reduction and more than 2× FPS improvement while preserving localization and mapping accuracy. These results highlight rendering-area–aware pruning as a promising direction for scaling 3DGS-SLAM to real-world autonomous driving scenarios. Our code is publicly available at https://github.com/UMN-ZhaoLab/Pocket-SLAM. git.