Wall Crack Detection Using Transfer Learning-based CNN Models
Sayyed Bashar Ali, Reshul Wate, Sameer Kujur, Anurag Singh, Santosh Kumar
Abstract
Early detection of cracks in building walls, roofs, bridges, etc. is quite important as these are early indicators for the ageing, decaying or any internal structural fault. This paper aims to develop an automatic inspection system based on deep learning model and image processing to identify cracks. Transfer-learning models of convolutional neural networks (CNNs) are used to learn the intrinsic features of cracks using the images of the surfaces, which help them for the automatic classification into cracked/un-cracked classes. We have explored, MobileNetV2, ResNet101, VGG16 and InceptionV2 architectures of CNN model and presented a performance comparison analysis of these models for crack detection in this work. After evaluating our proposed approach of crack-detection on publicly available datasets, we have found that out of all the pre-trained CNN models MobileNet yields the best performance with 99.59% detection accuracy after 10-fold cross validation and outperforms the state-of-the-art crack detection approaches.