Deep Learning Approaches for Autonomous Driving to Detect Traffic Signs

Traffic signs serve a vital function in regulating traffic flow, ensuring driver compliance with rules, and ultimately, enhancing road safety by reducing accidents and fatalities. The effective management of traffic signs, especially through automated Identification and acknowledgment are crucial elements within any Intelligent Transportation System (ITS). Given the advancements in self-driving vehicles, the demand for automated both automated detection and recognition of traffic signs has progressed increasingly imperative. This research study introduces an autonomous system that utilizes deep learning for the identification of traffic signs in India. The method for autonomously identifying and acknowledging traffic signs relies on the YOLOv5 framework, offering a comprehensive learning solution from start to finish. The proposed concept was evaluated using a novel benchmark known as the German Traffic Sign Recognition Benchmark. This benchmark comprises an extensive dataset containing greater than 50,001 visuals regarding roadway signage, categorized into 43 distinct classes or categories. Two-cascaded feature extraction method is employed which is a two-stage process, allowing the network to learn progressively more abstract and discriminative features, enhancing its Capability to apprehend intricate patterns. This method enhances the model's resilience to fluctuations in the dataset and is adaptable to specific tasks, reducing overfitting and potentially improving efficiency while maintaining interpretability. Furthermore, we carried out a performance assessment in contradiction to traditional Faster R-CNN and Mask R-CNN are two deep learning network configurations. Our suggested design demonstrated a remarkable training accuracy of 97.80% compared to other CNN methods.