Object Detection Methods for an Assistive System: A Survey to Improve Accessibility for the Visually Impaired

Author Name : Dr. Siyamol Chirakkarottu, Maanas. P. K, Rajasree. M, Sivada. K. M, Tania. M. Manoj

ABSTRACT

Visually impaired persons (VIPs) comprise a significant portion of the population, and they are present around the globe and in every part of the world. Vision is one of the foremost necessary senses that human beings use to interact with surrounding objects. This paper investigates the effectiveness of various object detection models for identifying objects in images for blind people. Four different models were considered: CNN, R-CNN, the Viola-Jones algorithm, and various versions of YOLO. After extensive testing, YOLO v5 was determined to be the best method for visually impaired object detection.

The paper provides a thorough examination of each model’s strengths and weaknesses in detecting objects in images. Accuracy, speed, and the   ability   to identify multiple objects simultaneously were among the evaluation criteria. In terms of accuracy and speed, YOLO v5 outperformed the other models, making it an excellent choice for real-time object detection.

Overall, the results of this study indicate that YOLO v5 is the best object detection model for the visually impaired. This research could be useful in developing assistive technologies for the visually impaired that rely on object detection, such as navigation aids or object recognition systems.