A comparison of deep convolutional neural networks for image-based detection of concrete surface cracks
The aim of this paper is to compare the performance of four deep convolutional neural networks in the problem of image-based automated detection of concrete surface cracks in the case of a small dataset. This crack detection problem is treated as a binary classification problem, and it is solved by training a deep convolutional neural network on the small dataset. In this context, overfitting during training was the main issue to cope with and various techniques were applied to overcome this issue. The results of the experiments suggest that the best approach for this problem is to use the pretrained convolutional base of a large pretrained convolutional neural network as an automatic feature extraction method and adding a new binary classifier on top of the convolutional base. Then, at the training the new classifier and fine-tuning the last few layers of the pretrained network take place at the same time. The classification accuracy of the best deep convolutional neural network on the testing set is about 94%.