Published Paper Details:

In today's world, visually impaired individuals encounter numerous challenges in daily activities such as recognizing objects, identifying currency denominations, and reading printed text. To address these difficulties, we propose a real-time Android-based assistive application that leverages deep learning models for independent navigation and interaction. The system integrates the YOLOv8 object detection algorithm for both general object and currency recognition, Optical Character Recognition (OCR) for text extraction, and Text- to-Speech (TTS) technology for audio feedback. Gesture-based swipe interactions enable users to seamlessly switch between functionalities, enhancing usability without requiring visual cues. Experimental evaluations demonstrate that the YOLOv8 model achieves an object detection accuracy of 94.5% and currency recognition accuracy of 95.2% across various lighting and environmental conditions. The OCR module yields a text extraction accuracy of 91.3% with minimal latency. The application achieves real-time processing speeds with an average inference time of 180 milliseconds per frame on a standard Android smartphone. By operating fully offline, the system ensures privacy, fast response, and accessibility. The proposed solution empowers visually impaired individuals to perform daily tasks independently, offering a practical, scalable, and highly effective assistive technology.