In this paper, we present a comparison of performance for different convolutional neural networks (CNN) for automaticclassification of corrosion and coating damages on bridge constructions from images. Image recordings were taken duringinspections. Through manual categorization and data augmentation, a total of 9300 images were collected and divided intofive classes. Four different CNNs were trained using transfer learning in MATLAB. We have evaluated test performancethrough the metrics recall, precision, accuracy and F1 score. Test performance was also evaluated on damage detectionaccuracy, meaning how well the networks detect images that contain a damage. The convolutional neural network trainedusing VGG-16 had the overall best performance results, with average recall, precision, accuracy and F1 score being95.45%, 95.61%, 97.74% and 95.53%, respectively. In the category of overall damage detection AlexNet performed bestwith 99.14% accuracy. The obtained results are promising, and make it possible to conclude that CNNs have a greatpotential in bridge inspections for automatic analysis of corrosion and coating damages.
展开▼
