Visually impaired individuals face significant challenges in navigating the world, and driving, a crucial aspect of independent living, becomes particularly daunting. While autonomous vehicle technology holds promise, its widespread adoption and affordability for all remain distant realities. Addressing this pressing need, we present ViziAssist, an innovative visual assistance system specifically designed to empower visually impaired drivers. Utilizing an NVIDIA Jetson Nano at its core and leveraging the power of the YOLOv7 object detection model, ViziAssist offers real-time object detection and classification capabilities tailored for road scenarios. This paper details the system’s architecture, the training methodology employing transfer learning on a custom dataset of Indian roads, and the implementation of an intuitive visual output mechanism using strategically placed LEDs. Rigorous testing procedures, encompassing both indoor and on-field evaluations, were conducted to assess the system’s performance. ViziAssist achieved an impressive mean Average Precision (mAP) of 0.681, demonstrating its efficacy in real-time object detection. We achieved particularly high precision scores for crucial object classes like autorickshaws (0.799) and buses (0.8), highlighting the system’s reliability in identifying potential hazards on Indian roads.