CNN Based on Transfer Learning Models Using Data Augmentation and Transformation for Detection of Concrete Crack

Abstract

Cracks in concrete cause initial structural damage to civil infrastructures such as buildings, bridges, and highways, which in turn causes further damage and is thus regarded as a serious safety concern. Early detection of it can assist in preventing further damage and can enable safety in advance by avoiding any possible accident caused while using those infrastructures. Machine learning-based detection is gaining favor over time-consuming classical detection approaches that can only fulfill the objective of early detection. To identify concrete surface cracks from images, this research developed a transfer learning approach (TL) based on Convolutional Neural Networks (CNN). This work employs the transfer learning strategy by leveraging four existing deep learning (DL) models named VGG16, ResNet18, DenseNet161, and AlexNet with pre-trained (trained on ImageNet) weights. To validate the performance of each model, four performance indicators are used: accuracy, recall, precision, and F1-score. Using the publicly available CCIC dataset, the suggested technique on AlexNet outperforms existing models with a testing accuracy of 99.90%, precision of 99.92%, recall of 99.80%, and F1-score of 99.86% for crack class. Our approach is further validated by using an external dataset, BWCI, available on Kaggle. Using BWCI, models VGG16, ResNet18, DenseNet161, and AlexNet achieved the accuracy of 99.90%, 99.60%, 99.80%, and 99.90% respectively. This proposed transfer learning-based method, which is based on the CNN method, is demonstrated to be more effective at detecting cracks in concrete structures and is also applicable to other detection tasks.

Public Database of Cracks Images in Mortar Coating with Different Types of Surface Finishes

“…This can be explained by the fact that the A1 group was already well-balanced, with a variation of 501 images between the classes in a universe of 22,795. The surface the smooth type would be the one that most resembled the surface of concrete; however, the results accuracies of A1 and B1 were 86.65% and 86.36%, respectively, which is below those of concrete studies like Chow et al [25], Ali et al [26], Islam et al [27], and Chaiysarn et al [28]. This may indicate that the application of computer vision in mortar coating images is actually more complex than in the case of concrete.…”

“…In the detection method proposed by the authors, they achieved 96.7% accuracy. Islam et al [27] developed a CNN approach based on transfer learning models using data augmentation and transformation for crack detection in concrete. The best research results were achieved using VGG16 and AlexNet, both with 99.9% accuracy.…”

A shallow 2D-CNN network for crack detection in concrete structures

“…To convincingly analyze the model's crack detection performances, we used a well-known dataset for implementation and execution (Islam et al ., 2022) as input images. For multi-class purposes, the primary data are divided into three classes.…”

BdSLW-11: Dataset of Bangladeshi sign language words for recognizing 11 daily useful BdSL words

“…Computer vision can greatly contribute to eradicating their suffering. Computer vision is being used for various real-life problems [3] . Researchers are developing various models or systems using computer vision to recognize hand signs and translate them into voice or text [4] .…”