AI City Challenge hits decade mark with 325 teams and 6 tracks

The 10th AI City Challenge, presented at ECCV 2026, celebrates a decade of community-driven benchmarking for intelligent transportation and smart cities. Since its 2017 inception with vehicle detection and tracking, the challenge has grown into a broad suite covering multi-camera perception, multimodal reasoning, synthetic-to-real learning, generative forecasting, and privacy-preserving evaluation. This year saw 325 registered teams—a 33% jump from 2025's 245—and participation expanded from 15 to 26 countries and regions.

The competition featured six primary tracks: multi-camera 3D perception, transportation safety captioning and VQA, traffic anomaly reasoning, text-based person anomaly search, generative traffic video forecasting, and cross-city object detection. Two additional out-of-domain leaderboards covered fisheye traffic-violation understanding and pedestrian situated-intent VQA. Successful solutions consistently combined foundation models with geometric grounding, retrieval or reranking, synthetic-data design, domain adaptation, and controlled inference. The paper details the datasets, evaluation protocols, and leaderboard results, serving as a comprehensive reference for researchers building real-world smart city AI.