Research map / Multi-agent systems and games / Human-AI and goal-conditioned
TopoMA: Topology-Guided Multi-Agent Dense RGB 3D Reconstruction via Distributed Inference
Xuanxuan Zhang, ShuHui Shi, Tianxiang Zhang, Zhetao Guo, Huang Zixuan, You Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add topology-aware loop closure (vs no loop closure) (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 23.81 -> 10.45 | Improved |
| use topology-aware loop closure instead of MASt3R-SLAM style loop closure (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 17.94 -> 10.45 | Improved |
| use topology-aware loop closure instead of ICP submap fusion (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 15.73 -> 10.45 | Improved |
| use multiple agents with loop closure instead of a single agent (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 11.68 -> 10.45 | Improved |
| add loop closure (compute cost) (ablation) | TopoMA · Replica apartment-00 | FPS (Hz): 7.35 -> 6.21 | Hurt |
| add topology-aware residual transport (vs fully independent agents) (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 18.34 -> 10.48 | Improved |
| use residual transport instead of centralized fusion (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 14.82 -> 10.48 | Improved |
| use residual transport instead of MNE-SLAM fusion (ablation) | TopoMA · Replica apartment-00 | ATE (cm): 16.71 -> 10.48 | Improved |
| fuse every frame-window lightly instead of coarse fusion every 500 frames (ablation) | TopoMA residual transport · Replica apartment-00 | ATE (cm): 12.32 -> 10.48 | Improved |
| use residual transport instead of centralized fusion (memory) (ablation) | TopoMA · Replica apartment-00 | GPU usage (GB): 6.50 -> 5.90 | Improved |
| TopoMA vs strongest single-agent baseline on KITTI-08 and KITTI-10 sequences | TopoMA · KITTI-08 / KITTI-10 | RMSE ATE (m): KITTI-10: 21.57 (VGGT-Long) vs 25.20 (Ours) | Hurt |
| TopoMA vs VGGT-Long on ScanNet reconstruction | TopoMA · ScanNet (5 scenes avg) | Avg. Depth L1 (cm): 60.83 -> 12.19 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MAGiC-SLAM — MAGiC-SLAM: Multi-Agent Gaussian Globally Consistent SLAM (CVPR 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MAC-Ego3D: Multi-Agent Gaussian Consensus for Real-Time Collaborative Ego-Motion and Photorealistic 3D Reconstruction (CVPR 2025 · similarity 0.88)
- FRUC: Feedforward Dynamic Scene Reconstruction from Uncalibrated Collaborative Driving Views (NeurIPS 2026 · similarity 0.85)
- MNE-SLAM: Multi-Agent Neural SLAM for Mobile Robots (CVPR 2025 · similarity 0.84)
- MAGiC-SLAM: Multi-Agent Gaussian Globally Consistent SLAM (CVPR 2025 · similarity 0.84)
- Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models (NeurIPS 2026 · similarity 0.84)
- 3DCodeBench: Benchmarking Agentic Procedural 3D Modeling Via Code (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
