Pothole-Road-Image

Pothole Road Image Classification and Segmentation
1.Pothole Road Image Classification：Pothole road detection and recognition is a computer vision task aimed at identifying roads with potholes by means of digital images (usually images of potholes in the ground). This is important for research and applications in fields such as geological exploration, aerospace science and natural disasters.A transfer learning based approach is used to learn the features of pothole images from a generic pre-trained model, EfficientNet, and use these features to classify pothole images.
2.Pothole Road Image Segmentation：The pothole image dataset is classified using the trained EfficientNet model, and then the pothole images are manually labelled with pothole edges using a labelling tool (labelme) and transformed into a mask map. Subsequently, the Dilated-UNet++ model is used for road pothole edge extraction and segmentation.