Automated Road Damage Detection Using UAV Images and Deep Learning Techniques

Abstract

This work presents an original strategy utilizing current deep learning techniques with "Unmanned Aerial Vehicle (UAV)" photographs to identify road decay consequently. Safe versatility relies upon well-keeping up with road foundation, however manual information gathering is in some cases hazardous and timeconsuming. We hence utilize "artificial intelligence (AI)” and UAVs to significantly further develop road damage recognizing accuracy and efficiency. For object identification in UAV photographs, our methodology utilizes “three state-of- the-art algorithms— YOLOv5 and YOLOv7”. YOLOv7 shows the best precision as per broad preparation and tests utilizing datasets from China and Spain. We further expand our examination by presenting YOLOv8, which shows essentially more forecast accuracy when prepared on road damage information than past frameworks. These outcomes feature the conceivable outcomes of “UAVs and deep learning in road damage distinguishing proof”, subsequently opening the way for next improvements in this area.