Deep Learning-Based Real-Time Object Detection, Recognition and Audio Assist System for Visually Challenged Persons

Blind or visually impaired people face numerous hurdles in their day today life and rely on the assistance of others. The majority of them rely on a cane or a dog for assistance to notice the obstacles in their path. To overcome these problems, deep learning algorithm ‘YOLO’ is used to identify and recognize the obstacles in the path, and these names are then converted to audio by using Pyttsx3 python package. In this system, speech command from user is taken to do particular process. In addition, we can direct the user to know whether the object is in his centre, right or left by considering the width of the image. Counting the number of objects present in front of user is done by count function. Ultrasonic sensors along with buzzers and micro-coin vibrators are implemented with Raspberry Pi to avoid obstacles by providing buzz sound and vibration. For reading the contents in image or printed documents, we do text recognition using Tesseract optical character recognition (OCR). The recognized text is then converted into audio output.