Challenge Winners 2026
Congratulations to the winners and runners-up of the 10th AI City Challenge, announced at the ECCV 2026 workshop on September 8, 2026.
The 2026 challenge included six tracks. The results below reflect the final code, eligibility, and reproducibility reviews completed by the track teams.
Track 1: Multi-Camera 3D Perception (Sim2Real)
Winner: Team 289 EVA
Runner-up: Team 34 SKKU-AL-T1
The following table shows the final verified results for Track 1.
| Rank | Team ID | Team Name | 3D HOTA (%) |
| 1 | 289 | EVA | 56.5447 |
| 2 | 34 | SKKU-AL-T1 | 48.2336 |
| 3 | 130 | Playbox | 38.0105 |
| 4 | 4 | VTX_QDTers | 34.1845 |
| 5 | 133 | TU-YMLab | 25.8837 |
Track 2: Transportation Safety Understanding and Captioning (Sim2Real)
Winner: Team 47 Latent Painter – UTE
Sim-to-Real Traffic Scene Understanding by Decoupling Semantics from Caption Generation with V-JEPA
Runner-up: Team 24 UIT – Kitchen
A Semantic State Bridge for Sim2Real Traffic-Safety VQA and Captioning
The following table shows the final verified results for Track 2.
| Rank | Team ID | Team Name | Score (S2) |
| 1 | 47 | Latent Painter – UTE | 60.0853 |
| 2 | 24 | UIT – Kitchen | 57.3307 |
| 3 | 266 | KZ6 | 56.7949 |
| 4 | 257 | MobilityAI | 55.5901 |
| 5 | 71 | Snow leopard | 55.5768 |
Track 3: Anomalous Events in Transportation
Traffic Anomaly Reasoning (TAR) Winner:: Team 25 Stellarview AI
Reason or Recite: Complementary Specialists with an LLM Judge for Traffic Anomaly Reasoning
Traffic Anomaly Reasoning (TAR) Runner-up: Team 45 UOB&UW Team
TAU-Agent: An Agentic Retrieval-Augmented Framework for Traffic Anomaly Understanding
The following table shows the final results for the TAR leaderboard.
Team 60 placed third on the final Public leaderboard and received an honorable mention without an award
| Final Rank | Team ID | Team Name | Mean Score |
| 1 | 25 | Stellarview AI | 0.6788 |
| 2 | 45 | UOB&UW Team | 0.6779 |
| 3 | 60 | FPT AI Vision | 0.6696 |
| 4 | 12 | Smart Vision | 0.6669 |
| 5 | 129 | CHT-IVS | 0.6563 |
FishEye Traffic Violation (FETV) Winner: Team 139 MR-CAS
The following table shows the final results for the FETV leaderboard. Note the FETV award order changed after final verification.
| Final Rank | Team ID | Team Name | Public Score | Verified Score |
| 1 | 139 | MR-CAS | 0.4884 | 0.4848 |
| 2 | 30 | UWIPL_ETRI | 0.4891 | 0.4629 |
| 3 | 277 | Korea Drive | 0.4634 | 0.4618 |
Egocentric Dashcam Pedestrian Intent (PSI VQA) Winner: Team 30 UWIPL_ETRI
The following table shows the final results for the PSI-VQA leaderboard.
| Final Rank | Team ID | Team Name | Mean Score |
| 1 | 30 | UWIPL_ETRI | 70.6397 |
| 2 | 257 | MobilityAI | 69.0698 |
| 3 | 122 | OptimAI | 65.5476 |
| 4 | 139 | MR-CAS | 64.4161 |
| 5 | 277 | Korea Drive | 57.0400 |
Track 4: Text-Based Person Re-Identification (Sim2Real)
Winner: Team 59 Xiilab.AIpex
Runner-up: Team 9 hiensumi
Protocol-Aware Global Assignment for Text-Based Person Anomaly Search
The following table shows the leaderboard and final award decisions for Track 4.
Eligibility note: Team 66 held public rank 2 but did not provide an eligible paper and code package. Verified Team 9 therefore received the runner-up award.
| Rank | Team ID | Team Name | mAP (%) |
| 1 | 59 | Xiilab.AIpex | 99.3020 |
| 2 | 66 | Hallucination | 99.2551 |
| 3 | 9 | hiensumi | 98.3535 |
| 4 | 137 | UIT-OpenCubee | 95.9370 |
| 5 | 103 | Safe AI | 95.9250 |
Track 5: Generative Traffic Video Forecasting
Winner: Team 209 Qyn
CosmosAlign: Adapting a World Foundation Model for Generative Traffic Video Forecasting
Runner-up: Team 78 SSUPER
The following table shows the final verified results for Track 5.
| Rank | Team ID | Team Name | mAP (%) |
| 1 | 209 | Qyn | 76.3917 |
| 2 | 78 | SSUPER | 76.0385 |
| 3 | 47 | Latent Painter – UTE | 75.1297 |
| 4 | 83 | CHTTL_A30 | 74.0544 |
| 5 | 39 | VGU_ai_lab | 73.2843 |
Track 6: Cross-City Object Detection (Milestone Project Hafnia)
Winner: Team 34 SKKU-AL-T1
Rethinking Pre-Training and Augmentation for Zero-Shot Cross-City Object Detection
Runner-up: Team 265 BIT-ODL
EC-DEIM: Evidence-Conditioned Multi-Source Pretraining for Cross-City Fine-Grained Object Detection
The following table shows the final verified results for Track 6.
| Rank | Team ID | Team Name | mAP |
| 1 | 34 | SKKU-AL-T1 | 0.4753 |
| 2 | 265 | BIT-ODL | 0.4281 |
| 3 | 152 | Remote Vibecoders | 0.4176 |
| 4 | 261 | BK2TheFuture | 0.4169 |
| 5 | 162 | S2 Detection | 0.4114 |
Congratulations to all award-winning teams!
Thank you to every team that participated in the 2026 AI City Challenge and to the track organizers who completed the final evaluations and reproducibility reviews.
For prize details and fulfillment requirements, please see the 2026 Challenge Awards page.