Research map / 3D vision / Depth and stereo
Debiased Multiplex Tokenizer for Efficient Map-Free Visual Relocalization
Wenshuai Wang, Hong Liu, Shengquan Li, Peifeng Jiang, Runwei Ding
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 |
|---|---|---|---|
| use ViM encoder instead of VGG16 (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.129 -> 0.073 | Improved |
| use ViM encoder instead of DINOv2-ViT (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.101 -> 0.073 | Improved |
| use SKSA instead of NLSA (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.103 -> 0.073 | Improved |
| add RBF convolution (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.097 -> 0.073 | Improved |
| use sparsemax activation (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.085 -> 0.073 | Improved |
| use TCDA instead of VLAD aggregation (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.144 -> 0.073 | Improved |
| use TCDA instead of GeM pooling (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.156 -> 0.073 | Improved |
| add spatial branch to TCDA (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.142 -> 0.073 | Improved |
| add frequency branch to TCDA (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.096 -> 0.073 | Improved |
| add channel branch to TCDA (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.129 -> 0.073 | Improved |
| use HNSW instead of kNN retrieval (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m) and per-token time: 0.076 -> 0.073; 64ms -> 5ms | No clear effect |
| use causal Mamba pointer CMP instead of Transformer decoder (ablation) | DeMT · 7Scenes (average translation error, m) | average translation error (m): 0.183 -> 0.073 | Improved |
Closest papers by meaning
- Local Supports Global: Deep Camera Relocalization With Sequence Enhancement (ICCV 2019 · similarity 0.86)
- Long-Term Visual Localization Revisited (TPAMI 2020 · similarity 0.86)
- VectorReLoc: Reliable Vectorized SD Map Visual Re-localization with Contrastive Feature Alignment (ECCV 2026 · similarity 0.85)
- Reloc3r: Large-Scale Training of Relative Camera Pose Regression for Generalizable, Fast, and Accurate Visual Localization (CVPR 2025 · similarity 0.84)
- RELOCATE: A Simple Training-Free Baseline for Visual Query Localization Using Region-Based Representations (CVPR 2025 · similarity 0.84)
- Change-Robust Online Topological Memory for Long-Term Relocalization and Semantic Navigation (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.
