Research map / Point clouds / Point cloud classification
VLM-Loc: Localization in Point Cloud Maps via Vision-Language Models
Shuhao Kang, Youqi Liao, Peijie Wang, Wenlong Liao, Qilin Zhang, Benjamin Busam, Xieyuanli Chen, Yun Liu
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 partial node assignment (PNA) on top of scene graph (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 24.62 -> 32.34 | Improved |
| add PNA to BEV + scene graph (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 29.79 -> 35.91 | Improved |
| add scene graph input to BEV-only input (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 13.21 -> 29.79 | Improved |
| add BEV image to scene-graph-only input (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 24.62 -> 29.79 | Improved |
| use partial instead of full node assignment (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 17.81 -> 35.91 | Improved |
| add directional cues to text query (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 18.01 -> 35.91 | Improved |
| add color cues to text query (with direction) (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 16.93 -> 18.01 (without direction) | No clear effect |
| scale backbone from Qwen3-VL-8B to 32B (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 35.91 -> 41.05 | Improved |
| scale backbone from Qwen3-VL-2B to 8B (ablation) | VLM-Loc · CityLoc-K test | Recall@5m: 34.70 -> 35.91 | No clear effect |
| replace CMMLoc with VLM-Loc | T2P localization · CityLoc-K test | Recall@5m: 21.71 -> 35.91 | Improved |
| transfer VLM-Loc to unseen CityLoc-C | T2P localization · CityLoc-C | Recall@5m: 13.68 (MNCL) -> 21.37 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CMMLoc — CMMLoc: Advancing Text-to-PointCloud Localization with Cauchy-Mixture-Model Based Framework (CVPR 2025) · better on 5 of 5 shared results
- Text2Loc — Text2Loc: 3D Point Cloud Localization from Natural Language (CVPR 2024) · better on 5 of 5 shared results
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
- Text2Loc: 3D Point Cloud Localization from Natural Language (CVPR 2024 · similarity 0.91)
- Recognition through Reasoning: Reinforcing Image Geo-localization with Large Vision-Language Models (NeurIPS 2025 · similarity 0.89)
- Text to Point Cloud Localization with Multi-Level Negative Contrastive Learning (AAAI 2025 · similarity 0.89)
- CMMLoc: Advancing Text-to-PointCloud Localization with Cauchy-Mixture-Model Based Framework (CVPR 2025 · similarity 0.89)
- GRE Suite: Geo-localization Inference via Fine-Tuned Vision-Language Models and Enhanced Reasoning Chains (NeurIPS 2025 · similarity 0.89)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.88)
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.
