Smart Sight: Real Time Scene Captioning for Visually Impaired

ABSTRACT: Visual impairment significantly limits an individual’s ability to perceive and interact with the surrounding environment, making independent navigation a daily challenge. To address this issue, this journal presents a vision-based assistive system designed to provide real-time environmental awareness and safety support for visually impaired users. The proposed system integrates a camera-based object detection module with distance sensing and audio feedback to convey meaningful information about nearby obstacles and objects. A pre-trained YOLO (You Only Look Once) deep learning model is employed for efficient and accurate object detection due to its high detection speed and capability to identify multiple objects in a single frame. To enhance safety during navigation, an ultrasonic sensor continuously measures the distance to nearby obstacles and triggers alerts when objects are detected within a predefined range. Detected information is communicated to the user through audio output, ensuring hands-free operation and minimizing cognitive load. In addition to environmental awareness, the system incorporates an emergency alert mechanism that allows the user to send their real-time location to a designated caregiver using a GPS and wireless communication module. This feature ensures immediate assistance during critical situations. The proposed system is designed to be portable, cost-effective, and user-friendly, making it suitable for real-world deployment. Experimental observations demonstrate that the system effectively improves situational awareness and enhances the independence and safety of visually impaired individuals. Keywords— Intelligent Assistive Vision System; Deep Learning–Based Object Detection; Proximity-Aware Obstacle Sensing; Context-Aware Audio Guidance; Emergency Location Broadcasting; Smart Accessibility Device.