Research map / Detection and segmentation / 3D detection and driving
Flow-Based Feature Fusion for Vehicle-Infrastructure Cooperative 3D Object Detection
Haibao Yu, Yingjuan Tang, Enze Xie, Jilei Mao, Ping Luo, Zaiqing Nie
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 FFNet feature-flow fusion instead of DiscoNet | DAIR-V2X cooperative detection at 200 ms · DAIR-V2X val, 200 ms | mAP@BEV, IoU=0.5: 58.20 -> 63.20 | Improved |
| Use FFNet feature-flow fusion instead of early fusion | DAIR-V2X cooperative detection at 300 ms · DAIR-V2X val, 300 ms | mAP@BEV, IoU=0.5: 58.28 -> 61.20 | Improved |
| Add feature prediction to compensate for latency (ablation) | FFNet · DAIR-V2X val, 200 ms | mAP@BEV, IoU=0.5: 57.93 -> 63.20 | Improved |
| Add feature prediction to compensate for latency (ablation) | FFNet · DAIR-V2X val, 0 ms | mAP@BEV, IoU=0.5: 63.54 -> 63.54 | No clear effect |
| Quantize feature flow to 6 bits (ablation) | FFNet · DAIR-V2X val, 200 ms | mAP@BEV, IoU=0.5: 63.20 -> 63.26 | No clear effect |
| Quantize feature flow to 6 bits (ablation) | FFNet · DAIR-V2X val, 300 ms | mAP@BEV, IoU=0.5: 61.20 -> 61.28 | No clear effect |
| Quantize feature flow to 6 bits (ablation) | FFNet · DAIR-V2X val, 200 ms | Average Byte (Byte): 1.2×10^5 -> 2.2×10^4 | Improved |
| Generate feature flow on infrastructure rather than on the vehicle (ablation) | FFNet versus FFNet-V (Same-TC) · DAIR-V2X val, 300 ms | mAP@BEV, IoU=0.5: 58.02 -> 61.20 | Improved |
| Train the feature-flow generator with additional unlabeled infrastructure sequences (ablation) | FFNet · DAIR-V2X val, latency 100–500 ms | mAP@BEV, IoU=0.5 | Improved |
Closest papers by meaning
- TransIFF: An Instance-Level Feature Fusion Framework for Vehicle-Infrastructure Cooperative 3D Detection with Transformers (ICCV 2023 · similarity 0.89)
- DAIR-V2X: A Large-Scale Dataset for Vehicle-Infrastructure Cooperative 3D Object Detection (CVPR 2022 · similarity 0.88)
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.86)
- TUMTraf V2X Cooperative Perception Dataset (CVPR 2024 · similarity 0.85)
- SparseAlign: a Fully Sparse Framework for Cooperative Object Detection (CVPR 2025 · similarity 0.84)
- 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.
