Research map / Vision-language and multimodal / Cross-modal learning
Modality-Invariant Visual Odometry for Embodied Vision
Marius Memmel, Roman Bachmann, Amir Zamir
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 explicit modality-invariance training to RGB-D VOT (ablation) | VOT-B RGB-D on Gibson-4+ validation · Gibson-4+ validation | success: 88.2 -> 92.6 | Improved |
| Train with modality-invariance so Depth can be dropped at test time (ablation) | RGB-D VOT · Gibson-4+ validation, Depth dropped | success: 26.1 -> 60.9 | Improved |
| Train with modality-invariance so RGB can be dropped at test time (ablation) | RGB-D VOT · Gibson-4+ validation, RGB dropped | success: 75.9 -> 91.0 | Improved |
| Use Depth rather than RGB as the sole input modality (ablation) | VOT on Gibson-4+ validation · Gibson-4+ validation | success: 59.3 -> 93.3 | Improved |
| Add multi-modal pre-training to RGB-D VOT without an action prior (ablation) | VOT-B on Gibson-4+ validation · Gibson-4+ validation | success: 47.3 -> 85.7 | Improved |
| Add an action prior to pre-trained RGB-D VOT (ablation) | VOT-B on Gibson-4+ validation · Gibson-4+ validation | success: 85.7 -> 88.2 | Improved |
| Add an action prior to pre-trained RGB VOT (ablation) | VOT-B on Gibson-4+ validation · Gibson-4+ validation | success: 54.5 -> 59.3 | Improved |
| Add an action prior to pre-trained Depth VOT (ablation) | VOT-B on Gibson-4+ validation · Gibson-4+ validation | success: 83.2 -> 93.3 | Improved |
| Use RGB-D instead of Depth-only with pre-training and an action prior (ablation) | VOT-B on Gibson-4+ validation · Gibson-4+ validation | success: 93.3 -> 88.2 | Hurt |
| Add RGB observations to Depth-only VOT (ablation) | VOT with modality-invariance training · Gibson-4+ validation | success: 93.3 -> 92.6 | Hurt |
| Train VOT using only RGB rather than Depth (ablation) | VOT on Gibson-4+ validation · Gibson-4+ validation | success: 93.3 -> 59.3 | Hurt |
| Replace the proposed approach with Zhao et al.'s unified baseline | PointGoal navigation on Gibson-4+ validation · Gibson-4+ validation | success: 92.6 -> 64.5 | Hurt |
Closest papers by meaning
- Reinforcement Learning Meets Visual Odometry (ECCV 2024 · similarity 0.89)
- The Surprising Effectiveness of Visual Odometry Techniques for Embodied PointGoal Navigation (ICCV 2021 · similarity 0.88)
- VOCA: Visual Odometry with Codec Awareness (ECCV 2026 · similarity 0.88)
- Efficient Deep Visual and Inertial Odometry with Adaptive Visual Modality Selection (ECCV 2022 · similarity 0.88)
- LEAP-VO: Long-term Effective Any Point Tracking for Visual Odometry (CVPR 2024 · similarity 0.86)
- ZeroVO: Visual Odometry with Minimal Assumptions (CVPR 2025 · 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.
