Fog-Aware Adaptive YOLO for Object Detection in Adverse Weather

Object detection in adverse weather conditions such as foggy environments is one of the main challenges in autonomous vehicles due to the significant reduction in visibility and performance of sensors. Although there are many publications to modify object detection in foggy environments, they are unable to manage both normal and foggy scenarios at the same time. In this paper, we propose a fog-aware adaptive YOLO algorithm for object detection in foggy environments. Our method first categorizes images into two groups based on their level of fogginess, normal and foggy, using a novel fog evaluator algorithm. In the next step, a standard YOLO algorithm is applied to normal images, while an image-adaptive YOLO algorithm is used for foggy images. Our approach provides a dynamic solution to evaluate the fog level of input images and adjust the detection algorithm accordingly, which can be applied in various real-world applications such as autonomous vehicles. Experimental results on the VOC dataset demonstrate the effectiveness of our approach in improving object detection performance in foggy conditions. The proposed method has a reasonable improvement in mean average precision compared to existing state-of-the-art methods in foggy weather conditions.