Research map / Vision-language and multimodal / Cross-modal learning
VMLoc: Variational Fusion For Learning-Based Multimodal Camera Localization
Kaichen Zhou, Changhao Chen, Bing Wang, Muhamad Risqi U. Saputra, Niki Trigoni, Andrew Markham
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 |
|---|---|---|---|
| fuse RGB and depth with variational PoE, attention, and importance weighting | camera localization · 7-Scenes average | position error: 0.20 m -> 0.19m | Improved |
| fuse RGB and depth with variational PoE, attention, and importance weighting | camera localization · 7-Scenes average | orientation error: 7.56^{\circ} -> 7.01^{\circ} | Improved |
| fuse RGB and lidar-derived depth with variational PoE, attention, and importance weighting | camera localization · Oxford RobotCar average | position error: 23.8 m -> 17.5m | Improved |
| fuse RGB and lidar-derived depth with variational PoE, attention, and importance weighting | camera localization · Oxford RobotCar average | orientation error: 8.19^{\circ} -> 4.92^{\circ} | Improved |
| add depth input to the RGB-only VMLoc (ablation) | VMLoc · Oxford RobotCar average | position error: 23.9 m -> 17.5m | Improved |
| add depth input to the RGB-only VMLoc (ablation) | VMLoc · Oxford RobotCar average | orientation error: 7.22^{\circ} -> 4.92^{\circ} | Improved |
| replace attention-based feature concatenation with PoE fusion and importance weighting (ablation) | VMLoc · Oxford RobotCar average | position error: 19.3 m -> 17.5m | Improved |
| replace attention-based feature concatenation with PoE fusion and importance weighting (ablation) | VMLoc · Oxford RobotCar average | orientation error: 5.35^{\circ} -> 4.92^{\circ} | Improved |
| add importance weighting and the proposed unbiased objective to MVAE-based PoE fusion (ablation) | VMLoc · Oxford RobotCar average | position error: 18.7 m -> 17.5m | Improved |
| add importance weighting and the proposed unbiased objective to MVAE-based PoE fusion (ablation) | VMLoc · Oxford RobotCar average | orientation error: 5.16^{\circ} -> 4.92^{\circ} | Improved |
| use full VMLoc rather than the ablated fusion variants (ablation) | camera localization · Oxford RobotCar, RGB corruption lvl=1 | position error: 229.1 m; 235.5m; 241.9m | Mixed |
| use full VMLoc rather than the ablated fusion variants (ablation) | camera localization · Oxford RobotCar, RGB corruption lvl=2 | position error: 484.7 m; 478.0 m; 464.7m | Improved |
Closest papers by meaning
- 6D Camera Relocalization in Ambiguous Scenes via Continuous Multimodal Inference (ECCV 2020 · similarity 0.87)
- FusionVAE: A Deep Hierarchical Variational Autoencoder for RGB Image Fusion (ECCV 2022 · similarity 0.87)
- OFVL-MS: Once for Visual Localization across Multiple Indoor Scenes (ICCV 2023 · similarity 0.86)
- On the Generalization Capacities of MLLMs for Spatial Intelligence (ICLR 2026 · similarity 0.84)
- Unity by Diversity: Improved Representation Learning for Multimodal VAEs (NeurIPS 2024 · similarity 0.84)
- DiMSOD: A Diffusion-Based Framework for Multi-Modal Salient Object Detection (AAAI 2025 · 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.
