Visually impaired individuals often face significant challenges, including collision with unexpected obstacles, lack of real-time spatial awareness and identifying safe paths. While the existing solutions alert users about the obstacle, they fail to address the gap of guiding them a safe path to navigate around the obstacle. The proposed system provides a real-time, lightweight navigation system that processes live video input to detect obstacles and offer audio guidance about the safe path, which is designed for adaptability across different hardware. The system uses fine-tuned YOLOv11 to detect anomalies and obstacles in the road and the bounding box information to estimate the proximity of the obstacles. A heuristic risk evaluation mechanism analyses the obstacle, proximity and risk factor to determine a safe path. The fine-tuned YOLOv11 achieved a reliable performance of mAP@50 of 0.8643 on the BDD100K dataset and 0.8949 on the RAD dataset. In real-world testing, the system successfully guided users with 95% accuracy, with minimal delays in complex scenarios. This cost-effective reliable solution enhances mobility and independence for visually impaired individuals in dynamic environments.