||Mobile Phone-based Real-time Dangerous Object Recognition for the Visually Impaired
||(Seungbin Kim) ; (Hartono Rodi) ; (Tshibang Patrick A Kalend) ; (Namkyun Baik) ; (KyooJae Shin)
|| Android; Object recognition; SSD; YOLO; Visually impaired; Real-time
||This paper proposes a dangerous object recognition Android application for the visually impaired using a smartphone. Among the one-stage and two-stage detectors, which are representative object recognition methods, the one-stage method is more suitable for real-time object recognition. Among them, more accurate object recognition is possible using the SSD network. In addition, the visually impaired cannot see the screen or the front properly in their daily life, so it informs them of dangerous situations in advance through sound. The system helps the visually impaired using a priority queue that notifies the visually impaired with a different sound according to a high-priority object. The experiment results of the proposed system are expected to bring positive results when applied to the smartphones of visually impaired people.