Crack identification from concrete structure images using deep transfer learning

• Published in: 2022 2nd International Conference on Artificial Intelligence and Signal Processing (AISP) pp. 1 - 7
• Main Authors: Syed, Amena Qadri, Jothi, J.Angel Arul, Anusree, K
• Format: Conference Proceeding
• Language: English
• Published: IEEE 12-02-2022
• Subjects: convolutional neural network; Crack detection; data augmentation; Deep learning; Feature extraction; Image segmentation; machine learning; Neural networks; Signal processing; Training; Transfer learning
• ISSN: 2640-5768
• DOI: 10.1109/AISP53593.2022.9760670

Early crack identification of civil structures is an essential task to prolong the life of the structures and to promise public safety. This research aims to develop an automated crack identification system using deep learning models and the SDNET2018 dataset. Image augmentation is applied to overcome the effect of unbalanced data. Deep pre-trained models like VGG16, InceptionV3, ResNet-50, ResNet-101 and ResNet-152 are trained and tested using the cracked and uncracked images of decks and pavements from the dataset. The experimental results show that the classification models obtained using transfer learning on the cracked and non-cracked pavement and deck image dataset have accuracy values of 70.59%, 60.31%71.93%, 75.40%, and 74.77% for VGG-16, Inception V3, ResNet-50, ResNet-101, and Resnet-152 pretrained models respectively.