This project focuses on traffic light classification using a custom Dual-Softmax Output Layer. The aim is to design and train a specialized model for traffic light classification, allowing for simultaneous prediction of both the traffic light state and pictogram.