Traffic Sign Recognition (TSR) is an essential component of Advanced Driver Assistance Systems (ADAS) and intelligent transportation. This paper presents a cost-effective IoT-based TSR system using an ESP32-CAM for image acquisition and a backend server for processing. Due to limited edge-device capability, images are transmitted via a Telegram Bot for remote inference using the YOLOv3 deep learning model trained on the GTSRB dataset. To enhance robustness under real-world conditions such as occlusion and varying illumination, preprocessing techniques like CLAHE and data augmentation are applied. The system returns annotated results through a Telegram interface and a local GUI. Experimental results demonstrate high accuracy and reliable performance, validating the effectiveness of the proposed approach. The system also shows strong performance under partially occluded conditions, improving real-world applicability. Furthermore, the proposed architecture ensures low-cost deployment and scalability for smart transportation systems. This work highlights the potential of integrating IoT with deep learning for practical and accessible traffic monitoring solutions.