Obstacle classifier for ADAS system using Deep Learning

Summary — This study presents the implementation of a convolutional neural network (CNN) designed for classifying images within a unique dataset consisting of three categories: pedestrian, infrastructure, and vehicle for an Advanced Driver Assistance System (ADAS) in an electric vehicle. Including preprocessing of the database images, the network achieves an accuracy of 99.83%, demonstrating its effectiveness and reliability in image classification. The results highlight the robustness of the architecture and the efficiency of the selected hyperparameters, suggesting its potential for practical applications in real-world scenarios.