Pedestrian and Vehicle Detection Performance Analysis of Faster R-CNN under Extreme Weather Conditions

Abstract

Vehicle safety systems now have a significant problem in reliably and accurately detecting objects in adverse weather due to the ongoing advancements in autonomous driving technology. The main objective of this work is to assess how well the Faster R-CNN model performs when it comes to identifying cars and pedestrians in inclement weather, including rain, snow, and fog. In order to learn more about the accuracy and resilience of the model, the researchers trained and tested it using a variety of datasets, including a range of harsh weather conditions. The model's performance significantly degrades in poor visibility and small object identification, according to the testing data. This decline in effectiveness demonstrates how severely inclement weather affects detecting systems. The paper also addresses potential enhancements, such as parameter fine-tuning, model optimization, and the application of increased data augmentation methods to boost robustness. Expanding the dataset, enhancing image quality in adverse situations, and researching more sophisticated model architectures will be the main goals of future studies in order to increase detection accuracy and robustness in practical autonomous driving scenarios.