Scale Anchoring via Implicit Semantics and Sequence-Consistent Loop Closure for VGGT-Based SLAM
Scale Anchoring via Implicit Semantics and Sequence-Consistent Loop Closure for VGGT-Based SLAM
This research extends VGGT-SLAM 2.0 with fixed-root scale anchoring and short-sequence-assisted direct loop connections to address scale drift in dense monocular reconstruction. The method uses learned implicit semantic-geometric cues to provide scale anchors while maintaining overlap chains, and employs consistency checks for translation-scale alignment without auxiliary loop submaps. Results show a reduction in mean Sim(3) ATE from 0.1018 m to 0.0451 m on nine TUM sequences compared to the baseline, with improvements across various ablation studies and comparisons on Clio scenes.
The paper presents a method to mitigate scale drift in VGGT-based SLAM by introducing fixed-root scale anchoring and sequence-consistent loop closure, reporting specific improvements in Absolute Trajectory Error (ATE) on TUM sequences.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
number of sequences compared
17
View original evidence
rame nodes, without an auxiliary loop submap. A 17-sequence comparison lowers mean Sim(3) ATE on nine TUM sequences from 0.1018 to 0.0451 m relative to the VGGT-SLAM 2.0 baseline. Anchoring improves all seven no-loop module ablations. On three Clio scenes, a fixed-loop comparison favors the direct connection pipeline over the tested auxiliary-submap pipeline. A separate candidate audit exposes additional rejections o
Open source S4
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
rame nodes, without an auxiliary loop submap. A 17-sequence comparison lowers mean Sim(3) ATE on nine TUM sequences from 0.1018 to 0.0451 m relative to the VGGT-SLAM 2.0 baseline. Anchoring improves all seven no-loop module ablations. On three Clio scenes, a fixed-loop comparison favors the direct connection pipeline over the tested auxiliary-submap pipeline. A separate candidate audit exposes additional rejections o
Open source S4
ALAD selects an image pair between the initial submap and each eligible new submap; joint VGGT inference then uses learned implicit semantic-geometric cues to supply a scale-only anchor while preserving the original overlap chain. For loop closure, neighboring image pairs support a consistency check and translation-scale alignment. Each accepted candidate contributes one robust Sim(3) factor between existing center-f
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org