Employing Real-Time Object Detection for Visually Impaired People

Visually impaired and blind people face several difficulties in their daily life. This was the primary motivation of this work as to create and assemble an object detector that can assist people with visual impairments using OpenCV and TensorFlow API on Raspberry Pi and provide an audio output for the detected objects using Espeak; Text-to-Speech Synthesizer. Single Shot Detector (SSD) model with MobileNet v2 has been employed to perform the detection with high accuracy and processing speed. The scripts are written in Python which utilizes the model to recognize the objects with boxes and provide class of the objects. The recognized image category is extracted and stored in a text file. The developed system provides aid to a visually impaired person for performing tasks independently using real-time object detection and identification technology. Developed system can successfully provide information about detected object in the form of an audio output to the visually impaired person.