Title: Hierarchically Decoupled Mixture-of-Experts for Robust Traffic Sign Recognition in Complex Driving Scenarios

Original: arXiv:2606.01822v1 Announce Type: new Abstract: Traffic sign detection is a fundamental component of environmental perception in autonomous driving and intelligent transportation systems. However, most existing detectors rely on static inference with globally shared parameters, limiting their ability to adapt to diverse and unstructured traffic scenarios. As a result, a single static model often struggles to simultaneously handle both clear near-range samples and challenging conditions such as distant small targets or adverse weather environments. To address this limitation, we propose CBDES MoE TSR, a hierarchically decoupled heterogeneous mixture-of-experts(MoE) framework for traffic sign recognition. The proposed framework departs from the conventional globally shared parameter paradigm by introducing a heterogeneous You Only Look Once (YOLO) expert pool together with a lightweight gating network, enabling an image-level dynamic routing mechanism. Based on the semantic characteristics of the input image, the gating module selectively activates the most suitable expert model from the expert pool, enabling a shift from fixed parameter fitting to on-demand dynamic representation. This design enhances feature extraction capability for specific scenarios while maintaining controlled inference overhead. Experimental results demonstrate that the proposed method achieves a remarkable balance between detection accuracy and efficiency on the composite traffic sign dataset. Specifically, our method attains an mAP50-95 of 76.8%, yielding a 2.3% improvement over the baseline method (74.5%) while simultaneously reducing computational overhead by approximately 39.4%. These findings robustly validate the effectiveness of the proposed approach.

Rewrite: Accurate traffic sign detection serves as a cornerstone for environmental awareness within intelligent transportation systems and autonomous vehicles. Nevertheless, the majority of current detection models depend on static inference processes utilizing globally shared parameters, which hinders their capacity to adjust to the varied and unstructured nature of real-world traffic conditions. Consequently, a solitary static model frequently fails to effectively manage both straightforward, close-range instances and difficult scenarios, such as identifying small targets at a distance or operating under poor weather conditions. To overcome these constraints, we introduce CBDES MoE TSR, a novel framework for traffic sign recognition that employs a hierarchically decoupled, heterogeneous mixture-of-experts (MoE) architecture. Moving away from the traditional approach of globally shared parameters, our framework incorporates a lightweight gating network alongside a heterogeneous pool of You Only Look Once (YOLO) experts. This setup facilitates a dynamic routing mechanism at the image level. By analyzing the semantic features of the input image, the gating module activates the most appropriate expert from the pool, thereby transitioning the system from static parameter fitting to dynamic, on-demand representation. This architectural choice improves feature extraction for specific contexts without significantly increasing inference costs. Evaluations on the composite traffic sign dataset reveal that our approach successfully balances precision and efficiency. The model achieves an mAP50-95 score of 76.8%, representing a 2.3% gain over the baseline’s 74.5%, while also cutting computational costs by roughly 39.4%. These outcomes strongly support the efficacy of the proposed methodology.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC