Design and Implementation of Real-time Object Detection for Blind using Convolutional Neural Network

In 2017 the Indonesian Ministry of Health estimated that there were 3.75 million blind and visually impaired people living in Indonesia. In this paper, we propose a device in the form of smart glasses as an object detection system and as a system for detecting barriers / obstacles in front of blind that equipped with earphone as speakers to convey object information to support and facilitate blind people's activities. The system is composed of a Pi Noir V2 camera is used for taking pictures of objects, ultrasonic sensors to determine the distance an object from user. Data processing is carried out on the Raspberry Pi used and further gives a sound-shaped output about the results of image processing and its position, through headphones. The detecting system used the MobileNet SSD v2 model and the Convolutional Neural Network method. The accuracy of system is verified through experiments. The effectiveness is confirmed through experiments with the optimal mAP and loss values determined by the model parameter testing process.