Road Damage Classification using SSD Mobilenet with Image Enhancement

Abstract

Road maintenance is required to maintain roads in good condition. Identifying the type of road damage is one of the steps in road maintenance. There has been research on the detection of types of road damage. However, there are still limitations, such as the fact that not all types of road damage can be detected with sufficient accuracy. In this paper, we utilized datasets from RDD2020. These datasets are open and can be used by anyone. This dataset is trained with the SSD Mobilenet algorithm through an image enhancement procedure that increases the image's smoothness and sharpness. Image enhancement process is performed to add feature extraction. SSD Mobilenet chosen because it has a high level of accuracy and speed, making it suitable for use in road damage detection systems. The addition of image enhancement can increase the model's accuracy, without image enhancement, model accuracy is 73%, while with image enhancement, it is 75%.