The process of evaluating pavement performance, conventionally performed by visual surveys, tends to be slow and inefficient in countries with extensive networks. The development of object detection algorithms and the popularization of smartphones open the possibility of more automated evaluation processes. This paper proposes the evaluation of the detection of vertical road signs and pavement defects from images produced by drivers using computer vision techniques. A collected set of YouTube videos produced by Brazilian drivers was used to train and validate a convolution neural network model. Results indicated an overall precision of 74.9 %, with observed detection deficiency only for longitudinal cracks and alligator cracking. Potholes, patches, and traffic signs are properly detected (precision between 73 and 95 %, depending on the detected object) for pavement management applications. Provided the diversity of data produced in videos and the overall results obtained in this research, the work herein indicates the possibility of massive citizen participation in the data collection process toward roadway quality.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.