Research map / Detection and segmentation / 3D detection and driving
TransIFF: An Instance-Level Feature Fusion Framework for Vehicle-Infrastructure Cooperative 3D Detection with Transformers
Ziming Chen, Yifeng Shi, Jinrang Jia
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 unified positional encoding in CDA module (ablation) | TransIFF · DAIR-V2X val | AP@0.5: 48.93 -> 54.73 | Improved |
| add filter module on both sides (ablation) | TransIFF · DAIR-V2X val | Comm volume / AP@0.5: comm 16.64 -> 12.39 (log2), AP@0.5 48.93 -> 48.97 | Improved |
| add query optimization (concat+MLP deduplication) in feature magnet (ablation) | TransIFF · DAIR-V2X val | AP@0.5: 57.44 -> 59.62 | Improved |
| use embedding positional encoding instead of concat (ablation) | CDA module · DAIR-V2X val | AP@0.5: 58.18 -> 59.62 | Improved |
| share MLP weights for position embedding between vehicle and infrastructure (ablation) | CDA module · DAIR-V2X val | AP@0.5: 59.62 -> 59.43 | No clear effect |
| use concat+MLP instead of average pooling for fusing overlapping queries (ablation) | Query Optimization · DAIR-V2X val | AP@0.5: 57.88 -> 59.62 | Improved |
| TransIFF vs Where2comm at highest bandwidth | TransIFF · DAIR-V2X val | AP@0.5: 63.71 -> 59.62 | Hurt |
| increase pose noise to 0.6 m/0.6 deg | TransIFF · DAIR-V2X val | 3D AP@0.5: 59.3 -> 58.2 | No clear effect |
Closest papers by meaning
- Flow-Based Feature Fusion for Vehicle-Infrastructure Cooperative 3D Object Detection (NeurIPS 2023 · similarity 0.89)
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.87)
- TUMTraf V2X Cooperative Perception Dataset (CVPR 2024 · similarity 0.86)
- IS-Fusion: Instance-Scene Collaborative Fusion for Multimodal 3D Object Detection (CVPR 2024 · similarity 0.85)
- SparseAlign: a Fully Sparse Framework for Cooperative Object Detection (CVPR 2025 · similarity 0.85)
- BEVSync: Asynchronous Data Alignment for Camera-based Vehicle-Infrastructure Cooperative Perception Under Uncertain Delays (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.
