Research map / Vision-language and multimodal / Robot action and navigation
Weakly-Supervised Multi-Granularity Map Learning for Vision-and-Language Navigation
Peihao Chen, Dongyu Ji, Kunyang Lin, Runhao Zeng, Thomas H. Li, Mingkui Tan, Chuang Gan
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 WS-MGMap instead of CM2 | VLN-CE navigation · VLN-CE Val-Seen | SR: 42.9 -> 46.9 | Improved |
| Use WS-MGMap instead of CM2 | VLN-CE navigation · VLN-CE Val-Unseen | SR: 34.3 -> 38.9 | Improved |
| Transfer the R2R-Habitat model without fine-tuning | WS-MGMap · RxR-Habitat English Val-Seen | SR: 9.0 -> 14.0 | Improved |
| Transfer the R2R-Habitat model without fine-tuning | WS-MGMap · RxR-Habitat English Val-Unseen | SR: 10.0 -> 15.0 | Improved |
| Add a multi-granularity map combining fine-grained and semantic information (ablation) | WS-MGMap · VLN-CE Val-Unseen | SR: 31.6 and 33.3 -> 38.9 | Improved |
| Add an explicit map instead of using raw RGB-D alone (ablation) | WS-MGMap · VLN-CE Val-Unseen | SR: 26.5 -> 38.9 | Improved |
| Add weakly supervised instruction-relevant object localization loss (ablation) | WS-MGMap · VLN-CE Val-Unseen | SR: 33.1 -> 38.9 | Improved |
| Use soft rather than hard localization ground truth (ablation) | WS-MGMap localization task · VLN-CE Val-Unseen | SR: 35.2 -> 38.9 | Improved |
| Use last-layer rather than first-layer U-Net features for the fine-grained map (ablation) | WS-MGMap · VLN-CE Val-Unseen | SR: 32.1 -> 38.9 | Improved |
| Use last-layer rather than classification-layer U-Net features for the fine-grained map (ablation) | WS-MGMap · VLN-CE Val-Unseen | SR: 33.3 -> 38.9 | Improved |
| Add semantic hallucination beyond the current field of view (ablation) | WS-MGMap · VLN-CE Val-Unseen | SR: 37.3 -> 38.9 | Improved |
| Use DAgger rather than imitation learning alone (ablation) | WS-MGMap training · VLN-CE Val-Unseen | SR: 24.6 -> 38.9 | Improved |
Closest papers by meaning
- Towards Learning a Generic Agent for Vision-and-Language Navigation via Pre-Training (CVPR 2020 · similarity 0.87)
- Imagine Before Go: Self-Supervised Generative Map for Object Goal Navigation (CVPR 2024 · similarity 0.86)
- Reasoning in Visual Navigation of End-to-end Trained Agents: A Dynamical Systems Approach (CVPR 2025 · similarity 0.86)
- Hierarchical Semantic-Augmented Navigation: Optimal Transport and Graph-Driven Reasoning for Vision-Language Navigation (NeurIPS 2025 · similarity 0.86)
- NavGPT: Explicit Reasoning in Vision-and-Language Navigation with Large Language Models (AAAI 2024 · similarity 0.86)
- Learning with a Mole: Transferable latent spatial representations for navigation without reconstruction (ICLR 2024 · similarity 0.85)
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.
