Diseño de un prototipo de detección y reconocimiento señalético de semáforo peatonal para personas no videntes haciendo uso de visión artificial y sensores ultrasónicos

Abstract

This work presents a prototype of signal recognition of pedestrian traffic lights for blind people using artificial vision and ultrasonic sensors. The design, training and result of the CNN from the YOLOv4-tiny model on a Raspberry Pi 4 are shown, including the designs of the case and the obstacle detection system.