Analysis of Different Deep Learning Algorithms for Road Surface Damage Detection

Numerous asphalt pavement faults are the major contributor to auto accidents, necessitating corrective action because they put people in grave danger. As a result, there are many algorithms used to detect those road damages so that no further accidents occur in the future. A model is proposed which consist of Convolutional Neural Network and ResNet algorithm to find the accuracy in both sections. First, the training dataset is collected from the RDD2020 dataset, which consists of 7000 images of three different countries then labeling of those images, is done in different categories of cracks like longitudinal, alligator, potholes, and traverse cracks. Furthermore, we implement CNN and ResNet architecture to analyze the accuracy and use a better algorithm to detect road damage in the future. After applying the CNN and ResNet-34, 94.79% and 89.94% accuracies are obtained as an outcome.