Visual blockage assessment at culverts using synthetic images to mitigate blockage-originated floods

Abstract The assessment of visual blockages in cross-drainage hydraulic structures, such as culverts and bridges, is crucial for ensuring their efficient functioning preventing flash flooding incidents. extraction blockage-related information through computer vision algorithms can provide valuable insights into the blockage. However, absence comprehensive datasets has posed a significant challenge effectively training models. In this study, we explore use synthetic data, images culvert (SIC) hydraulics lab dataset (VHD), combination with limited real-world dataset, openings blockage (ICOB), to evaluate performance opening detector. Faster Region-based Convolutional Neural Network (Faster R-CNN) model ResNet50 backbone was used impact data evaluated two experiments. first involved different combinations while second reduced images. results experiment revealed that structured training, where were initial ICOB fine-tuning, resulted slightly improved detection performance. showed conjunction number images, significantly degradation rates.