Deep Learning-Based Object Detection, Face Recognition, and Tracking Support Model for Visually Challenged

The visually impaired people (VIPs) constitute a substantial section of the world’s population. VIPs struggle with performing daily chores and independent mobility in everyday life. The proposed approach aims to create a model that can detect objects in real time from their surroundings and translate them into speech that VIPs can hear. The model is further programmed to detect familiar faces and recognize the emotions of people around them, thus helping VIPs break barriers of social interaction and communication. Additionally, the model also aims at providing the user's real-time location to his peers and family to ensure the user's safety. The proposed model uses the convolutional neural network architecture and you only look once (YOLO) object detection technique to obtain low computational complexity that enables it to run efficiently on low-power devices. The entire system was assembled employing the Raspberry Pi 3, accompanied by the integration of a GSM module within the proposed configuration.