Smart Road Damage Detection

Author Name : Prof. Dr. Kamini Nalawade, Chetan Lahase, Suvita Mhasade, Krishna Tambe, Suyog Bhujbal

ABSTRACT

Road damages have resulted in numerous fatalities, research into the detection of road damage, particularly the identification and warning of hazardous road damage, is essential for traffic safety. The majority of data processing for existing road damage detection systems occurs in the cloud, which has a high latency from long-distance transmission. Meanwhile, supervised machine learning techniques are often applied in these systems that need sizable, meticulously labelled datasets to attain excellent performance. In this work, we advocate the utilisation of Deep Learning to identify and forewarn about road deterioration. Visual observations made by individuals and quantitative analysis performed using expensive technologies are the basis of road surface study. For instance, visual inspection requires time and money in addition to the use of skilled road managers. Visual inspection also carries a higher risk because it is inherently unreliable and inconsistent. Due to the length of roads or motorways, visual assessment of roads by engineers takes a lot of time. Setting up an AI-based automated system that can identify the type of damage can therefore help to enhance and improve how road conditions are assessed.

Index terms- Road Damage, image classification, Dataset, Safe driving experience, Deep Learning.