Deep Learning Virtual Assistant for Visually Impaired with Object Detection and Distance Estimation

According to WHO (World Health Organization) 285 million individuals worldwide suffer from vision impairment. These metrics may triple in the coming 30 years. Thus, there is a need to develop a cost-effective guiding system for those people. This system provided a solution using large dataset COCO (Common Objects in Context) dataset which consists of 90 classes of real-time objects to cover all the day-to-day life objects It used the Single Shot Detector algorithm for object detection. It makes the work of blind people easier, and more efficient by transferring wireless voice-based feedback about whether an object is too close to them or at a safer distance. Google's text-to-speech engine is being used to provide audio-based output. It estimates the distance from person to object using depth extraction methodology with focal length as a scaling factor, real object width. This model provides a guiding system to help visually impaired people