AI City Challenge 2026: Frozen detector achieves high cross-city object detection AP

A study for the AI City Challenge 2026 explored cross-city object detection using a frozen RF-DETR-Large detector. The research found that inferring at a higher resolution (1120x1120) with frozen parameters, trained at 704x704, yielded the best aggregate AP score among tested configurations. This approach showed significant gains, particularly for small and medium-sized objects, without any parameter updates. AI

IMPACT This research demonstrates a method for improving object detection generalization across different urban environments without retraining.

RANK_REASON Academic paper detailing a study for a specific challenge. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• AI City Challenge 2026
• COCO-style AP

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →