Identify Faults in Road Structure Zones with Deep Learning

Traffic accidents occur due to the deterioration of road structures, this poor condition can lie in the low quality of its components, climatic changes, seismic zones, heavy cargo transport traffic, among others; the purpose of this research is the detection of pavement faults. The research was carried out taking into account the following stages: collection and preparation of data that consisted of the capture of images with a smartphone in the city of Lima-Peru, resizing and labeling, architecture network where the training process was captured, processing that consists of the entry of the algorithm and the deployment that is the implementation of the same in a system to detect cracks and gaps; for which the YOLOv5 algorithm was used, where for the training of the algorithm 420 images with 100 epochs were used, having an accuracy of 58%, and then validated with 30 images, having the averages of 85% accuracy, 91.5 sensitivity and 81.5% F1-Score. It is concluded that the algorithm with convolutional neural networks helps to properly identify pavement failures so that the authorities can do the proper maintenance.