AI Based Object Detection for Blind Assistance

Abstract

Visually impaired individuals face significant challenges in identifying obstacles, navigating unfamiliar places, recognizing nearby objects, and accessing real-time information. Current assistance systems usually depend on costly sensors or lack smart automation. To fill this gap, we propose an Enhanced Real-Time Blind Assistance System. This system integrates live object detection, voice-command interaction, multilingual support, text-reading capability, and emergency alert handling. The system uses TensorFlow.js with the COCO-SSD model to detect objects through a live camera feed. It provides spoken feedback about object proximity and direction. Additionally, it monitors critical conditions and sends emergency notifications when necessary. This system operates entirely through a browser, so it requires no special hardware. This makes it easy to access and deploy. The findings indicate that the system detects objects accurately, delivers timely alerts, and responds effectively to voice commands. Ultimately, it improves mobility, safety, and helping visually impaired individuals move around on their own whether the user is indoor or outside.