A Novel Deep Learning Based Object Detection And Recognition Framework For The Visually-Impaired

Abstract

Blindness is one of the top ten disabilities in humans, and India unfortunately has the world's largest visually impaired population. In this paper, we present a novel framework for assisting the visually impaired in object detection and recognition, allowing them to navigate and be aware of their surroundings independently. The paper employs a yolo mechanism for object detection and classification, followed by human face and currency note recognition.Yolo detector is trained on the coco dataset, and a new class is added to detect currency as well. Finally, the framework's output can be presented to the visually impaired person in audio format. The mean Accuracy and Precision (mAP) scores of the standalone yolo detector of the added currency class were 67.8 percent, and the testing accuracy of person and currency recognition was 90 to 92 percent.