YOLOv12-based algorithm for foggy weather vehicle and pedestrian detection

The foggy weather vehicle and pedestrian detection algorithm proposed in this paper, based on YOLOv12, effectively captures edge textures of blurred targets in foggy conditions through a regional attention mechanism. By introducing a residual efficient layer aggregation network, it optimizes gradient propagation via residual connections to enhance discrimination capability for vehicles and pedestrians in low-contrast environments, thereby improving model accuracy. The algorithm employs rapid attention to reduce computational redundancy in self-attention mechanisms, lowering model complexity. Experimental results demonstrate that compared to YOLOv11, the proposed algorithm achieves a 2.1% improvement in mAP and a 10.5% enhancement in FPS, effectively ensuring foggy road traffic safety.