BlindEye: Blind Assistance Using Deep Learning

Blind Community faces significant challenges in independently navigating their surroundings, often relying on white canes or assistance from others. These traditional methods can limit their mobility and autonomy. Therefore, there is a pressing need for innovative solutions that leverage emerging technologies to enhance the navigation capabilities of blind individuals. Deep learning techniques have shown great promise in various computer vision tasks, making them a compelling approach to addressing the complexities of blind navigation. In this study, we are proposing a unified approach that provides blind assistance using obstacle avoidance, terrain awareness and person identification. We utilised U-FCHarDNet model and achieved the state-of-the-art mIOU of 67.82. With a processing time of only 20 ms, our approach is well-suited for real-time environments. By leveraging these advancements, we aim to provide efficient navigation assistance and support to the blind community.