This paper proposes a customized convolutional neural network for crack detection in concrete structures. The proposed method is compared to four existing deep learning methods based on training data size, data heterogeneity, network complexity, and the number of epochs. The performance of the proposed convolutional neural network (CNN) model is evaluated and compared to pretrained networks, i.e., the VGG-16, VGG-19, ResNet-50, and Inception V3 models, on eight datasets of different sizes, created from two public datasets. For each model, the evaluation considered computational time, crack localization results, and classification measures, e.g., accuracy, precision, recall, and F1-score. Experimental results demonstrated that training data size and heterogeneity among data samples significantly affect model performance. All models demonstrated promising performance on a limited number of diverse training data; however, increasing the training data size and reducing diversity reduced generalization performance, and led to overfitting. The proposed customized CNN and VGG-16 models outperformed the other methods in terms of classification, localization, and computational time on a small amount of data, and the results indicate that these two models demonstrate superior crack detection and localization for concrete structures.

Chest radiography is a significant diagnostic tool used to detect diseases afflicting the chest. The automatic detection techniques associated with computer vision are being adopted in medical imaging research. Over the last decade, several remarkable advancements have been made in the field of medical diagnostics with the application of deep learning techniques. Various automated systems have been proposed for the rapid detection of pneumonia from chest X-rays. Although several algorithms are currently available for pneumonia detection, a detailed review summarizing the literature and offering guidelines for medical practitioners is lacking. This study will help practitioners to select the most effective and efficient methods from a real-time perspective, review the available datasets, and understand the results obtained in this domain. It will also present an overview of the literature on intelligent pneumonia identification from chest X-rays. The usability, goodness factors, and computational complexities of the algorithms employed for intelligent pneumonia identification are analyzed. Additionally, this study discusses the quality, usability, and size of the available chest X-ray datasets and techniques for coping with unbalanced datasets. A detailed comparison of the available studies reveals that the majority of the applied datasets are highly unbalanced and limited, providing unreliable results and rendering methods that are unsuitable for large-scale use. Large-scale balanced datasets can be obtained via smart techniques, such as generative adversarial networks. Current literature has indicated that deep learning-based algorithms achieve the best results for pneumonia classification with an accuracy of 98.7%, a sensitivity of 0.99, and a specificity of 0.98. The higher accuracy offered by deep-learning algorithms in addition to their appropriate class balancing techniques serves as a good reference for further research.

The use of deep learning (DL) in civil inspection, especially in crack detection, has increased over the past years to ensure long-term structural safety and integrity. To achieve a better understanding of the research work on crack detection using DL approaches, this paper aims to provide a bibliometric analysis and review of the current literature on DL-based crack detection published between 2010 and 2022. The search from Web of Science (WoS) and Scopus, two widely accepted bibliographic databases, resulted in 165 articles published in top journals and conferences, showing the rapid increase in publications in this area since 2018. The evolution and state-of-the-art approaches to crack detection using deep learning are reviewed and analyzed based on datasets, network architecture, domain, and performance of each study. Overall, this review article stands as a reference for researchers working in the field of crack detection using deep learning techniques to achieve optimal precision and computational efficiency performance in light of electing the most effective combination of dataset characteristics and network architecture for each domain. Finally, the challenges, gaps, and future directions are provided to researchers to explore various solutions pertaining to (a) automatic recognition of crack type and severity, (b) dataset availability and suitability, (c) efficient data preprocessing techniques, (d) automatic labeling approaches for crack detection, (e) parameter tuning and optimization, (f) using 3D images and data fusion, (g) real-time crack detection, and (h) increasing segmentation accuracy at the pixel level.

Missing values are common in real-world datasets and pose a significant challenge to the performance of statistical and machine learning models. Generally, missing values are imputed using statistical methods, such as the mean, median, mode, or machine learning approaches. These approaches are limited to either numerical or categorical data. Imputation in mixed datasets that contain both numerical and categorical attributes is challenging and has received little attention. Machine learning-based imputation algorithms usually require a large amount of training data. However, obtaining such data is difficult. Furthermore, no considerate work has been conducted in the literature that focuses on the effects of the training and testing size with an increasing amount of missing data. To address this gap, we propose that increasing the amount of training data will improve the imputation performance. We first used generative adversarial network (GAN) methods to increase the amount of training data. We considered two state-of-the-art GANs (tabular and conditional tabular) to add synthetic samples using observed data with different synthetic sample ratios. We then used three state-of-the-art imputation models that can handle mixed data: MissForest, multivariate imputation by chained equations, and denoising auto encoder (DAE). We propose a robust experimental setup on four publicly available datasets with different training-testing data divisions that have increasing missingness ratios. Extensive experimental results show that incorporating synthetic samples with training data achieves better performance compared to the baseline methods for mixed data imputation in both categorical and numerical variables, especially for large missingness ratios.

Recently, deep-learning (DL)-based crack-detection systems have proven to be the method of choice for image processing-based inspection systems. However, human-like generalization remains challenging, owing to a wide variety of factors such as crack type and size. Additionally, because of their localized receptive fields, CNNs have a high false-detection rate and perform poorly when attempting to capture the relevant areas of an image. This study aims to propose a vision-transformer-based crack-detection framework that treats image data as a succession of small patches, to retrieve global contextual information (GCI) through self-attention (SA) methods, and which addresses the CNNs’ problem of inductive biases, including the locally constrained receptive-fields and translation-invariance. The vision-transformer (ViT) classifier was tested to enhance crack classification, localization, and segmentation performance by blending with a sliding-window and tubularity-flow-field (TuFF) algorithm. Firstly, the ViT framework was trained on a custom dataset consisting of 45K images with 224 × 224 pixels resolution, and achieved accuracy, precision, recall, and F1 scores of 0.960, 0.971, 0.950, and 0.960, respectively. Secondly, the trained ViT was integrated with the sliding-window (SW) approach, to obtain a crack-localization map from large images. The SW-based ViT classifier was then merged with the TuFF algorithm, to acquire efficient crack-mapping by suppressing the unwanted regions in the last step. The robustness and adaptability of the proposed integrated-architecture were tested on new data acquired under different conditions and which were not utilized during the training and validation of the model. The proposed ViT-architecture performance was evaluated and compared with that of various state-of-the-art (SOTA) deep-learning approaches. The experimental results show that ViT equipped with a sliding-window and the TuFF algorithm can enhance real-world crack classification, localization, and segmentation performance.

Low Birth weight (LBW) infants pose a serious public health concern worldwide in both the short and long term for infants and their mothers. Infant weight prediction prior to birth can help to identify risk factors and reduce the risk of infant morbidity and mortality. Although many Machine Learning (ML) algorithms have been proposed for LBW prediction using maternal features and produced considerable model performance, their performance needs to be improved so that they can be adapted in real-world clinical settings. Existing algorithms used for LBW classification often fail to capture structural information from the tabular dataset of patients with different complications. Therefore, to improve the LBW classification performance, we propose a solution by transforming the tabular data into a knowledge graph with the aim that patients from the same class (normal or LBW) exhibit similar patterns in the graphs. To achieve this, several features related to each node are extracted such as node embedding using node2vec algorithm, node degree, node similarity, nearest neighbors, etc. Our method is evaluated on a real-life dataset obtained from a large cohort study in the United Arab Emirates which contains data from 3453 patients. Multiple experiments were performed using the seven most commonly used ML models on the original dataset, graph features, and a combination of features, respectively. Experimental results show that our proposed method achieved the best performance with an area under the curve of 0.834 which is over 6% improvement compared to using the original risk factors without transforming them into knowledge graphs. Furthermore, we provide the clinical relevance of the proposed model that are important for the model to be adapted in clinical settings.