Pot Hole Detection Using Deep Learning

Abstract

Potholes can holes on the outermost layer of roads that measure more than 75 millimeters horizontally and 20 millimeters deep. They are caused by overloading cars, poor structure, blocked water during the rainy season, decay of rocks, and occasionally by all of these factors together. According to statistics, 58,208 collisions caused more over 57,000 injuries and over 57,000 fatalities in our nation during the past 20 years. And potholes are a big factor in many of these incidents. Due to this pothole, cyclists are in serious risk today. To identify potholes at any time and warn vehicles to avoid any discomfort or accidents, a reliable detection apparatus is required. Although considerable work has been done on this issue, we provide a method that will meet the demands for real-time crater detection using deep learning.  On the attributes of our amassed dataset, we tried a few deep learning algorithms and obtained encouraging results. Additionally, our model's ability to spot potholes in the moment will help save many lives.