Influence of different factors for the Accura...

Influence of different factors for the Accuracy in the Pothole Detection System

Author Name : Ashish Kumar Barai

ABSTRACT

According to the latest official figures given in the Rajya Sabha, in 2015 people lost their lives in 4,869 road accidents caused by potholes while 4,108 people were injured.In 2017, 3,597 people lost their lives in the country due to road accidents caused by potholes. The most populous state Uttar Pradesh accounted for the maximum number of deaths that year.Potholes were one of the leading causes of accidents in 2019, accounting for nearly 4,770 incidents.Despite the fact that the number of incidents caused by potholes has dropped, their proportion in the total number of accidents caused by road conditions has grown. Pothole has been a big factor for the deaths of lives since a very long time.In this study, we use Deep Learning to detect whether a given image of road is a pothole or not, by feeding the images of road to the trained Convolution Neural Network model. The model has been trained on a dataset containing images of normal roads and roads with potholes. It was observed that most of the previous models did not focus much on the preprocessing of the dataset as a result of which the accuracy doesn’t go beyond 80 percent but when proper preprocessing of the dataset is done the accuracy can go beyond 95 percent which is a good increase in accuracy.

Keywords: Accuracy, Deep Learning, Convolution Neural Network, Pothole detection, Preprocessing