Safe Autonomous Driving in Rainy Weather Conditions using Deep Learning

Abstract

Rainy weather conditions have a substantial impact on the efficiency of autonomous driving systems, disrupting their ability to perceive the surrounding environment accurately. It can cause distortion in the input of the sensors, making it challenging for image-based perception systems to perform their tasks such as object detection and localization. This can result in degraded performance and potentially unsafe driving in adverse conditions. To address this challenge, we propose a deep learning-based approach enabling autonomous vehicles to navigate roads in rainy weather conditions safely. By using a combination of image processing techniques and deep learning algorithms, we aim to improve the robustness, accuracy, and reliability of autonomous driving systems in challenging weather conditions. Our approach has the potential to significantly enhance the safety and effectiveness of autonomous vehicles, making them more suitable for real-world deployment. We used the Rain13k dataset, which is synthetic rain dataset, to train a HINet-based deep learning model with enhancements. As a part of the enhancement, we increased the depth of the HIN Block in a certain manner, which enabled us to obtain a much better performance of the HINet model. We also reduced the number of iterations by 10 times. Our deep learning-based approach shows promising results in enabling safe autonomous driving in rainy weather conditions and improving their robustness. We also generated a paired image dataset containing real rain steaks in order to test the performance of the models in real-world scenarios and proposed a model which performed better on real-world data as compared to existing models. Future work could focus on incorporating other environmental factors such as snow, fog, and haze to further improve the performance of autonomous driving systems. There is an immense need for a larger dataset of real rain so that models can perform well in real-world scenarios. There are still a lot of challenges that need to be resolved in order to make autonomous vehicles much safer.