M^3: Dense Matching Meets Multi-View Foundation Models for Monocular Gaussian Splatting SLAM
This work addresses the problem of high-precision pose estimation and efficient online refinement in dynamic environments for SLAM applications, representing an incremental improvement by enhancing existing multi-view foundation models with dedicated matching capabilities.
The paper tackles the challenge of streaming reconstruction from uncalibrated monocular video by developing M^3, which integrates a multi-view foundation model with a matching head for dense correspondences into a monocular Gaussian splatting SLAM system, achieving state-of-the-art accuracy with a 64.3% reduction in ATE RMSE compared to VGGT-SLAM 2.0 and a 2.11 dB improvement in PSNR over ARTDECO on ScanNet++.
Streaming reconstruction from uncalibrated monocular video remains challenging, as it requires both high-precision pose estimation and computationally efficient online refinement in dynamic environments. While coupling 3D foundation models with SLAM frameworks is a promising paradigm, a critical bottleneck persists: most multi-view foundation models estimate poses in a feed-forward manner, yielding pixel-level correspondences that lack the requisite precision for rigorous geometric optimization. To address this, we present M^3, which augments the Multi-view foundation model with a dedicated Matching head to facilitate fine-grained dense correspondences and integrates it into a robust Monocular Gaussian Splatting SLAM. M^3 further enhances tracking stability by incorporating dynamic area suppression and cross-inference intrinsic alignment. Extensive experiments on diverse indoor and outdoor benchmarks demonstrate state-of-the-art accuracy in both pose estimation and scene reconstruction. Notably, M^3 reduces ATE RMSE by 64.3% compared to VGGT-SLAM 2.0 and outperforms ARTDECO by 2.11 dB in PSNR on the ScanNet++ dataset.