The 10th AI City Challenge

Details

The AI City Challenge, launched in 2017 with basic vehicle detection, has grown into one of the largest annual computer vision competitions focused on intelligent transportation and smart city AI. The 10th edition, held with ECCV 2026, attracted 325 registered teams from 26 countries — up from 245 teams and 15 countries in 2025 — marking steady growth in community engagement. The six primary tracks for 2026 cover: multi-camera 3D perception, transportation safety captioning and visual question answering, traffic anomaly reasoning, text-based person anomaly search, generative traffic video forecasting, and cross-city object detection. Two additional out-of-domain leaderboards (Tracks 7 and 8) covered fisheye traffic-violation understanding and pedestrian situated-intent VQA. Across all tracks, successful systems in 2026 share a common pattern: foundation models grounded with geometric or domain-specific priors, combined with retrieval/reranking, synthetic data augmentation, and domain adaptation. The inclusion of generative traffic video forecasting as a track signals the field moving from perception toward predictive world models for transportation.