Benchmark RadarOmniTraffic
OmniTraffic is a controllable generation pipeline and benchmark for spatio-temporal traffic reasoning. It provides 8M VQA samples and a 3K human-verified test set across 12 reconstructed 3D intersections, with a three-level task hierarchy covering scene perception, multi-view and temporal reasoning, and decision support.
- Released
- 2026-06-14
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
OmniTraffic fills the gap in evaluating structure-aware traffic reasoning under controlled conditions, where existing traffic benchmarks focus on passive recognition. The extensible pipeline allows configurable scenarios for reproducible evaluation and simulation-generated supervision for model fine-tuning.
Motivation
Traffic scene understanding requires models to reason beyond object recognition, including lane topology, multi-view geometry, temporal evolution, and signal-phase semantics.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.