FogNet: An Enhanced Object Detection Model for Vehicle and Human Recognition in Foggy Conditions

Abstract : Foggy weather makes it really hard for vehicle detection systems to work properly. The visibility drops, and objects on the road become hard to recognize. To tackle this problem, we developed a smart and lightweight detection method based on an improved version of the YOLOv10 model. This system doesn't just rely on raw images-it first applies a series of advanced preprocessing techniques. These include data transformations, Dehaze Formers, and dark channel methods that help clean up the foggy images and bring out the important details. By doing this, we reduce the effect of haze and make the key features more visible. We also added a special attention module to the model. This helps the system focus better on the important parts of the image by understanding both the surroundings and finer details. It's especially useful for spotting small or partially hidden vehicles and people in foggy scenes. On top of that, we improved the feature extraction process using a lightweight yet powerful module. This makes the system faster and more efficient, without compromising on accuracy. Overall, our approach offers a solid and reliable solution for detecting vehicles and humans even in tough foggy conditions, making roads safer and detection systems more dependable. Keywords: Foggy Weather, Vehicle Detection, Human Detection, YOLOv10, Lightweight Model, Image Preprocessing, Fog Removal, Dehaze Techniques, Dark Channel Method, Attention Module, Feature Extraction, Object Detection, Low Visibility, Real-Time Detection, Deep Learning, Computer Vision, Road Safety, Smart Detection System, Adverse Weather Detection, Autonomous Driving.