An application of Faster R-CNN for the detection and recognition of Ecuadorian traffic signs

This paper presents an application of Faster R-CNN in the development of a software for traffic signs recognition; i.e, this work implements an object detector based on Faster-RCNN with ZF-Net. The entire training and testing were developed on a database taken in urban environments from several cities in Ecuador. This dataset consists of 52 classes, collected in the various lighting environments (dawn, day, sunset and cloudy) from 6 am to 7 pm. After that, several experiments were carried out in real road driving conditions by using a technology platform, which consists of a vehicle for the implementation of driving assistance systems using Computer Vision and Artificial Intelligence.