VisionAid: A visual support tool

Visual impairment poses significant challenges to independent living by limiting perception, navigation, and interaction with the environment. To address these challenges, this study proposes VisionAid, a real-time assistive system that integrates YOLOv5-based object detection with text-to-speech (TTS) for audio feedback. The system leverages deep learning models fine-tuned on diverse datasets, combined with preprocessing and augmentation techniques, to ensure robust performance in varied indoor and outdoor settings. Implemented using Python, PyTorch, and OpenCV, VisionAid delivers accurate object recognition with a mean Average Precision (mAP) of 89.2% and real-time inference speed of 0.12 seconds per frame, making it suitable for deployment on edge devices. User trials with visually impaired participants highlighted improvements in awareness, confidence, and navigation. While limitations such as reduced low-light accuracy and interference from ambient noise were observed, planned enhancements include OCR integration, gesture/voice commands, scene-level understanding, and multilingual support. Overall, VisionAid demonstrates a scalable, efficient, and user-centric approach to enhancing autonomy and quality of life for visually impaired individuals.