Real-time regional road sign detection and identification using Raspberry Pi

Abstract

In India, there often arises a situation where mobile mapping/navigation services doesn’t always show the updated routes to the destination with accounting of frequent regional changes, that for an emergency vehicle may lead up to a dangerous situation, this can be resolved by using real-time image processing system in complementary with mapping/navigation services to recognize the street name sign boards and display it to the driver. In this paper, an affordable regional road sign detection and identification system is developed which can be fitted in vehicles for driver awareness. EfficientDet-Lite model architecture and TensorFlow Lite API is used for object identification in real-time through implementing it into a Raspberry Pi 4b board. To assess the performance of the system, various test was undertaken and the result show 100% detection precision and an average precision over 78% has been achieved with low latency.