Pothole Detection Using Jetson Nano Embedded System – An Evaluation of Training Models

The field of “smart city” has recently seen an increasing demand on the technology consumer market, this leads to the need to develop new solutions to meet the demands of users. A smart city is defined as an urban area that uses various types of electronic methods and sensors to collect data. Statistics obtained from this data are used to manage assets, resources, and services effectively; moreover, this data is used to improve or optimize the various activities within the city. The present research consisted of studying and implementing different neural network training models using the NVIDIA Jetson Nano development kit. This is a small embedded system capable of running Artificial Intelligence applications supported by the parallel processing capabilities of the embedded GPU. Using this built-in system, certain real-time applications can be developed for image classification or object detection, in our case asphalt pits.