Autonomous vehicles (AV) face a significant challenge in navigating rainy conditions due to the visual impairment of camera-based systems by rain droplets. This research endeavor aims to address this issue by harnessing contemporary deep learning methods. The primary objective is to develop a model that can process live images from vehicle cameras, eliminating rain-induced visual hindrances and yielding visuals closely resembling clear, rain-free images. A simulation environment in CARLA is utilized to capture a diverse range of clear and rainy images for training and testing. The model’s architecture is devised based on the designs of Deep Convolution Generative Adversarial Network and U-Net, and is trained with a novel batching scheme to enhance dynamic rain features in successive image frames. A predictor module is integrated to demonstrate the benefits of deraining in improving steering angle prediction, thereby enhancing AV performance and safety in adverse weather.