Research map / Detection and segmentation / 3D detection and driving
DAIR-V2X: A Large-Scale Dataset for Vehicle-Infrastructure Cooperative 3D Object Detection
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Fuse vehicle and infrastructure predictions with late fusion | PointPillars · VIC-Sync, DAIR-V2X-C | AP3D overall: 31.33 -> 41.90 | Helped |
| Fuse vehicle and infrastructure predictions with late fusion | PointPillars · VIC-Sync, DAIR-V2X-C | AP3D overall: 17.62 -> 41.90 | Helped |
| Fuse raw vehicle and infrastructure point clouds with early fusion instead of late fusion | PointPillars · VIC-Sync, DAIR-V2X-C | AP3D overall: 41.90 -> 50.03 | Helped |
| Use early fusion of raw point clouds instead of late fusion | PointPillars · VIC-Sync, DAIR-V2X-C | AB (Byte): 336.16 -> 1382275.75 | Hurt |
| Replace synchronous pairs with one-frame-earlier infrastructure data | PointPillars late fusion · VIC-Async-1, DAIR-V2X-C | AP3D overall: 41.90 -> 40.21 | Hurt |
| Replace synchronous pairs with two-frames-earlier infrastructure data | PointPillars late fusion · VIC-Async-2, DAIR-V2X-C | AP3D overall: 41.90 -> 35.29 | Hurt |
| Estimate infrastructure-object motion and compensate predictions to the vehicle timestamp before late fusion | PointPillars late fusion · VIC-Async-1, DAIR-V2X-C | AP3D overall: 40.21 -> 40.79 | Helped |
| Estimate infrastructure-object motion and compensate predictions to the vehicle timestamp before late fusion | PointPillars late fusion · VIC-Async-2, DAIR-V2X-C | AP3D overall: 35.29 -> 36.72 | Helped |
| Use image-based late fusion with vehicle and infrastructure views | ImvoxelNet · VIC-Sync, DAIR-V2X-C | AP3D overall: 12.03 -> 26.56 | Helped |
| Use SECOND instead of PointPillars | DAIR-V2X-V single-view detection · DAIR-V2X-V | Vehicle 3D AP, Easy: 61.76 -> 69.44 | Helped |
| Use MVXNet instead of SECOND | DAIR-V2X-V single-view detection · DAIR-V2X-V | Pedestrian 3D AP, Easy: 43.45 -> 47.73 | Helped |
| Use MVXNet instead of SECOND | DAIR-V2X-I single-view detection · DAIR-V2X-I | Pedestrian 3D AP, Easy: 55.16 -> 55.83 | Helped |
Closest papers by meaning
- DI-V2X: Learning Domain-Invariant Representation for Vehicle-Infrastructure Collaborative 3D Object Detection (AAAI 2024 · similarity 0.89)
- TUMTraf V2X Cooperative Perception Dataset (CVPR 2024 · similarity 0.89)
- BEVSync: Asynchronous Data Alignment for Camera-based Vehicle-Infrastructure Cooperative Perception Under Uncertain Delays (AAAI 2025 · similarity 0.87)
- UrbanIng-V2X: A Large-Scale Multi-Vehicle, Multi-Infrastructure Dataset Across Multiple Intersections for Cooperative Perception (NeurIPS 2025 · similarity 0.87)
- HoloVIC: Large-scale Dataset and Benchmark for Multi-Sensor Holographic Intersection and Vehicle-Infrastructure Cooperative (CVPR 2024 · similarity 0.87)
- AGC-Drive: A Large-Scale Dataset for Real-World Aerial-Ground Collaboration in Driving Scenarios (NeurIPS 2025 · 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.