Image-based Multi-Damage Detection in Tunnels: a Deep Learning dataset for Structural Health Monitoring

Abstract: The paper presents a dataset for image-based damage detection in tunnels and compares multiple models in detecting these anomalies. It introduces first the dataset and explaines how it was collected and processed to label observable damages. It then benchmarks Mask RCNN, Cascade Mask RCNN and Mask2Former in the instance segmentation performance.