Research map / 3D vision / 3D generation
LangLoc: “Tell Me What You See”
Shaurya Kishore Panwar, Roham Zendehdel Nobari, Shirley Lau, Abu Bakr Rahman Shaik, Manuel Günther, Marc Pollefeys, Daniel Barath
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 Bayesian yes/no dialog (ablation) | LangLoc · 3RScan (100 scenes) | median position error (m): 1.55 -> 0.80 | Improved |
| add Bayesian yes/no dialog (ablation) | LangLoc · 3RScan (100 scenes) | mean angular error (deg): 46.1 -> 39.5 | Improved |
| add Bayesian yes/no dialog (ablation) | LangLoc · 3RScan (100 scenes) | 3D IoU: 0.172 -> 0.342 | Improved |
| add Bayesian yes/no dialog (ablation) | LangLoc · ScanNet (100 scenes) | median position error (m): 1.31 -> 0.07 | Improved |
| add Bayesian yes/no dialog (ablation) | LangLoc · ScanNet (100 scenes) | median angular error (deg): 34.7 -> 5.1 | Improved |
| add Bayesian yes/no dialog (ablation) | LangLoc · ScanNet (100 scenes) | 3D IoU: 0.236 -> 0.593 | Improved |
| add human-typed dialog answers (ablation) | LangLoc · Human dialog pilot (10 scenes) | mean position error (m): 1.74 -> 0.66 | Improved |
| use ray-cast visibility scoring instead of zero-shot Qwen2.5-VL-2B | LangLoc · 3RScan (100 scenes) | mean angular error (deg): 85.5 -> 46.1 | Improved |
| use LangLoc without dialog instead of floor-centroid midpoint | LangLoc w/o dialog · 3RScan (100 scenes) | mean position error (m): 1.416 -> 1.712 | Hurt |
| use LangLoc dual-branch GATv2 retrieval instead of Text2SGM (ret-based) | LangLoc · ScanScribe retrieval (10-scene pool) | Top-1 recall (%): 68.61 -> 76.70 | Improved |
| use LangLoc instead of Text2SGM (cos-sim) on full 55-scene pool | LangLoc · ScanScribe retrieval (55-scene pool) | Top-5 recall (%): 76.34 -> 83.30 | Improved |
| use LangLoc instead of Text2SGM under LLM-generated queries | LangLoc · ScanScribe retrieval (LLM-generated queries) | Top-1 recall (%): 34.22 -> 59.5 | Improved |
Closest papers by meaning
- InLoc: Indoor Visual Localization With Dense Matching and View Synthesis (CVPR 2018 · similarity 0.87)
- LangMap: A Human-Verified Benchmark for Hierarchical Open-Vocabulary Goal Navigation (NeurIPS 2026 · similarity 0.85)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.84)
- LOCATE 3D: Real-World Object Localization via Self-Supervised Learning in 3D (ICML 2025 · similarity 0.84)
- Rooms from Motion: Un-posed Indoor 3D Object Detection as Localization and Mapping (NeurIPS 2025 · similarity 0.84)
- 360Loc: A Dataset and Benchmark for Omnidirectional Visual Localization with Cross-device Queries (CVPR 2024 · 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.
