Mitigating Transportation Risks: Addressing Safety Challenges in Inclement Weather Conditions

Frequent accidents resulting from adverse weather conditions worldwide have prompted the need for improved safety measures. Tragic events like automobile disasters, train derailments, airline crashes, and ship collisions have been featured in recent headlines. This paper presents deep convolutional neural networks (CNNs), namely Sequential, VGG, and ResNet50 models, for weather condition recognition by utilizing the progress made in deep learning. The dataset used in this study encompasses real-life data collected from Kaggle. The dataset comprises images representing five distinct weather classes, such as Shine, Rainy, cloudy, foggy, and sunrise, emphasizing the practical challenges that any weather classification system must address. With approximately 1500 labelled images, including validation data, the dataset reflects the variability in image dimensions and sizes. Based on the specified class, each photograph is arranged into distinct folders and corresponds to a specific weather category. The analysis of numerous experiments shows that the suggested ResNet50 outperforms Sequential and VGG algorithms in terms of recognition accuracy. This research on deep learning-based weather condition recognition using CNNs addresses the urgent need for enhanced safety measures in transportation, aiming to mitigate frequent accidents caused by adverse weather conditions worldwide, thereby contributing to societal well-being and safety. Moreover, the societal impact extends beyond transportation safety, influencing various sectors dependent on weather-sensitive activities such as agriculture, emergency response, and urban planning.