Towards Road Anomaly Detection for Autonomous Vehicle System Using Deep Learning Technique

Safety on road is still the top priority, despite the development of intelligent as well as autonomous car technologies brought about by recent advances in automobile technology. A move towards vision-based sensors and deep learning approaches for better scene interpretation has been prompted by the high costs and patchy performance of conventional sensors like radar and LiDAR for the safety of vehicles. It is crucial to detect road irregularities accurately, like potholes and speed bumps Convolutional Neural Networks (CNNs) are optimized for improved anomaly detection in this research study, the accuracy and performance of the model have been improved by carefully regulating the batch size and experimenting with various optimizers a dataset of 1900 labeled photos was used to test the CNN architecture which was modeled around LeNet-5 the adam optimizer was used for fine-tuning resulting in a detection accuracy of 94.2% and using common assessment matrices, the suggested model's performance is verified. This optimized approach enhances safety and reliability in autonomous driving systems.