Machine Learning for Visually Impaired: A Comprehensive Review of Object Detection Models in Assistive Technologies

Abstract

Visual impairment affects over 285 million individuals worldwide, significantly limiting their independence and interaction with their environments. In recent years, machine learning has emerged as a pivotal enabler in developing assistive technologies that address these challenges, particularly through object detection models. This comprehensive review evaluates the effectiveness of leading object detection algorithms—YOLO, SSD, and Faster R-CNN—within assistive applications such as wearable smart glasses, mobile apps, and robotic aids. Each model presents a unique balance between speed, accuracy, and computational efficiency, influencing their deployment across different platforms. While YOLO offers real- time responsiveness, SSD provides a balanced trade-off between precision and latency, and Faster R-CNN delivers high accuracy at the expense of real-time performance. The paper discusses real-world implementation challenges, including lighting variability, model bias, false detections, and hardware constraints, and highlights the importance of user-centered evaluation metrics. Additionally, ethical considerations such as data privacy, inclusivity, and equitable access are explored. Future directions focus on hybrid modeling, edge AI optimization, and multimodal feedback systems to further enhance user autonomy. This review underscores the transformative potential of machine learning-driven object detection in im- proving the quality of life for visually impaired individuals while calling for more inclusive, efficient, and context-aware technology development.

Keywords: Machine Learning, R-CNN, Leading Object Detection Algorithms.