Search papers, labs, and topics across Lattice.
The 10th AI City Challenge, held alongside ECCV 2026, has evolved into a comprehensive benchmarking platform for intelligent transportation and smart cities, featuring six primary tracks that encompass multi-camera perception, multimodal reasoning, and generative forecasting. With participation increasing to 325 teams from 26 countries, the challenge highlights advancements in areas such as traffic anomaly reasoning and privacy-preserving evaluation. Key findings reveal that successful systems leverage a combination of foundation models, geometric grounding, and innovative synthetic-data strategies to enhance performance across diverse tasks.
The latest AI City Challenge reveals that integrating foundation models with synthetic data design can significantly boost performance in complex urban scenarios.
The 10th AI City Challenge, held with ECCV 2026, marks a decade of community benchmarking for intelligent transportation, smart cities, and physical AI. Since its 2017 start with vehicle detection, classification, and tracking, the challenge has grown into a broad benchmark suite for multi-camera perception, multimodal reasoning, synthetic-to-real learning, generative forecasting, and privacy-preserving evaluation. The 2026 edition continued this growth with 325 registered teams, up from 245 in 2025, and participation from 26 countries and regions, up from 15. Its six primary tracks cover 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. Track 3 further includes two out-of-domain leaderboards, submitted as Tracks 7 and 8, for fisheye traffic-violation understanding and pedestrian situated-intent VQA. This paper summarizes the challenge setup, datasets, evaluation protocols, leaderboard results, and workshop papers. Across tracks, successful systems combine foundation models with geometric grounding, retrieval or reranking, synthetic-data design, domain adaptation, and controlled inference.
