Research map / Vision-language and multimodal / Cross-modal learning
Linking Modality Isolation in Heterogeneous Collaborative Perception
Changxing Liu, Zichen Chao, Siheng Chen
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 |
|---|---|---|---|
| modality isolation of m1 and m6 in training data (vs non-isolated) (ablation) | Pyramid Fusion · OPV2V m1 + m6 | AP70: 80.88 -> 65.67 | Hurt |
| add codebook to end-to-end training (ablation) | Pyramid Fusion · OPV2V m1 + m6 group | AP70: 80.44 -> 77.87 | Hurt |
| freeze encoder and add adapter after codebook (ablation) | CodeAlign · OPV2V m1 + m6 group | AP30/AP70: AP30 89.54 -> 88.37; AP70 77.87 -> 78.09 | Mixed |
| add similarity loss L_sim (ablation) | CodeAlign group code space · OPV2V m1 + m6 group | AP70: 78.09 -> 79.63 | Improved |
| use Instance similarity loss instead of none (ablation) | CodeAlign group code space · OPV2V m1 + m6 group | AP70: 78.09 -> 76.08 | Hurt |
| use group code space construction instead of FCF translation for non-isolated modalities (ablation) | CodeAlign · OPV2V m1 + m6 | AP70: 72.92 -> 79.63 | Improved |
| translate dense-to-code instead of code-to-code (ablation) | CodeAlign translator · OPV2V m1 + m6 | AP70: 55.16 -> 73.47 | Improved |
| translate dense-to-code instead of dense-to-dense (ablation) | CodeAlign translator · OPV2V m1 + m6 | AP70: 75.04 -> 73.47 | Hurt |
| use multi-head translator instead of one-to-one (ablation) | CodeAlign · OPV2V 4 isolated modalities | AP50: 88.61 -> 88.51 | No clear effect |
| increase codebook size from 4 to 16 (ablation) | CodeAlign · OPV2V m1 + m6 | AP70: 66.30 -> 73.47 | Improved |
| increase codebook size from 16 to 64 (ablation) | CodeAlign · OPV2V m1 + m6 | AP70: 73.47 -> 74.56 | No clear effect |
| train translator on single-agent data instead of collaborative data (ablation) | CodeAlign · OPV2V m1 + m2 | AP50: 94.88 -> 94.24 | No clear effect |
Closest papers by meaning
- NegoCollab: A Common Representation Negotiation Approach for Heterogeneous Collaborative Perception (NeurIPS 2025 · similarity 0.90)
- An Extensible Framework for Open Heterogeneous Collaborative Perception (ICLR 2024 · similarity 0.88)
- What Makes Good Collaborative Views? Contrastive Mutual Information Maximization for Multi-Agent Perception (AAAI 2024 · similarity 0.88)
- Strong Helps Weak: Directional Cross-Modal Alignment Transfer in Multi-modal LLMs (NeurIPS 2026 · similarity 0.87)
- DeepAlign: Mitigating Modality Conflict through Modality-Specific Alignment (CVPR 2026 · similarity 0.87)
- When to Align, When to Predict: A Phase Diagram for Multimodal Learning (NeurIPS 2026 · similarity 0.87)
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.
