Authors: Miss. Mounika Lokavarapu, Dr.G. Sharmila Sujatha
Abstract: Object recognition plays a crucial role in computer vision applications, particularly in assisting visually impaired individuals for safe and independent navigation. Despite its significance, existing techniques often face limitations in recognizing multiple objects efficiently and accurately. The aim of this work is to develop a robust multi-label object recognition framework capable of detecting and classifying surrounding objects in real time to enhance situational awareness for visually impaired users. The proposed system takes real-world images as input and processes them using machine learning and advanced computer vision algorithms. A multi-label classification approach is employed to simultaneously detect and group objects, reducing detection time while improving recognition accuracy. By leveraging deep learning models with optimized type/grouping techniques, the system achieves faster execution with best-in-class time complexity. Experimental analysis demonstrates that the framework not only improves detection performance but also provides reliable object recognition in both indoor and outdoor environments, making it highly effective for real-world navigation assistance. The proposed framework, “AI-Powered Assistive Vision: A Novel Deep Learning Framework for Object Detection and Recognition for the Visually Impaired,” is developed using Python with TensorFlow/Keras and OpenCV libraries, and implemented under embedded hardware with camera and processing units, enabling real-time deployment for assistive navigation applications.
DOI: http://doi.org/10.5281/zenodo.17098418
