Convolution neural networks for pothole detection of critical road infrastructure

A well developed and maintained highway infrastructure is essential for the economic social prosperity of modern societies. Highway maintenance poses significant challenges pertaining to ever-increasing ongoing traffic, insufficient budget allocations lack resources. Road potholes detection timely repair a major contributing factor sustaining safe resilient critical road infrastructure. Current pothole methods require laborious manual inspection roads in terms accuracy inference speed. This paper proposes novel application Convolutional Neural Networks on accelerometer data detection. Data collected using an iOS smartphone installed dashboard car, running dedicated application. The experimental results show that proposed CNN approach has advantage over existing solutions, with respect computational complexity