Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces DRAFE, a novel approach that combines LW-DETR and RF-DETR detectors to enhance fine-grained traffic object detection across different cities. By leveraging a two-stage training strategy that incorporates diverse datasets and human-in-the-loop refinement, DRAFE effectively addresses the challenges of domain generalization in traffic monitoring. The method achieves a competitive performance with a mean Average Precision (mAP) of 0.4022, ranking sixth in the AI City Challenge 2026 Track 6, and surpassing a preliminary ensemble by 0.0553 mAP.
DRAFE's innovative fusion of heterogeneous detection transformers significantly boosts cross-city traffic object detection accuracy, setting a new benchmark in the field.
Deep learning-based object detectors are fundamental to intelligent transportation systems, enabling traffic monitoring, vehicle analytics, and infrastructure management. However, achieving both fine-grained vehicle recognition and robust cross-city domain generalization remains challenging. We present the Domain-Robust Asymmetric Fusion Ensemble (DRAFE), which combines independently trained LW-DETR and RF-DETR detectors for cross-city fine-grained traffic object detection. DRAFE employs a two-stage training strategy that first pretrains complementary detectors on diverse public traffic datasets using pseudo-label expansion and human-in-the-loop annotation refinement, producing a curated corpus of 6,049 images and 203,619 annotations, before challenge-compliant fine-tuning on the Project Hafnia Track 6 dataset. At inference, DRAFE applies anchor-conditioned class-consistent matching, reliability-weighted coordinate fusion, agreement-aware confidence recalibration, and complementary hypothesis recovery. On AI City Challenge 2026 Track 6, DRAFE achieves 0.4022 mAP, ranks sixth among 25 participating teams, and improves by 0.0553 mAP over a preliminary ensemble evaluated under identical benchmark conditions.