Research on Real -Time Object Detection with Speech Feedback for Visually Impaired

Abstract: It is widely recognized that approximately 285 million people worldwide are visually impaired, which accounts for nearly 20% of India's population. Many of these individuals rely heavily on others to fulfill even their most basic daily needs. In our project, we utilized TensorFlow, a cutting-edge library developed by Google, to power our neural network models. The TensorFlow Object Detection API was employed to detect various objects in real-time. Additionally, we introduced the YOLO algorithm, which plays a pivotal role in object detection. During training, YOLO matches anchor boxes with the bounding boxes of ground truth objects in an image. The anchor box with the highest overlap with an object is responsible for predicting that object's class and location. Our system also integrates a microcontroller equipped with a Wi-Fi module for seamless communication. This user-friendly interface allows visually impaired users to navigate both indoor and outdoor environments with ease. To further enhance the user experience and reduce navigation challenges, we incorporated an obstacle detection system using ultrasonic sensors. This system identifies nearby obstacles and alerts the user, helping them avoid potential hazards and navigate with greater confidence.