In this paper, we propose a new methodology for crack detection and monitoring in concrete structures. This approach is based on a multiresolution analysis of a sample or a specimen of concrete material subjected to several types of solicitation. The image obtained by ultrasonic investigation and processed by a customized wavelet is analyzed at various scales in order to detect internal cracks and crack initiation. The ultimate objective of this work is to propose an automatic crack type identification scheme based on convolutional neural networks (CNN). In this context, crack propagation can be monitored without access to the concrete surface and the goal is to detect cracks before they are visible. This is achieved through the combination of two major data analysis tools which are wavelets and deep learning. This original procedure is shown to yield a high accuracy close to 90%. In order to evaluate the performance of the proposed CNN architectures, we also used an open access database, SDNET2018, for the automatic detection of external cracks.

On 16 March 2020, as a result of the unprecedented global health crisis linked to the emergence of a new form of coronavirus (COVID-19), the 74 universities of France closed their doors, forcing nearly 1.6 million students, as well as their teachers, to find solutions and initiatives that could ensure continuity in teaching. In the reliance on videoconferencing tools, chat, the sharing of documents/tutorials/videos/podcasts, and the use of social networks, many ideas have emerged, but no consensus has developed nor has a common way of doing things been adopted by a majority of teachers. Some software tools, such as Zoom, have also been questioned over data security issues or excessive intrusion into the student learning process. Nevertheless, in these uncertain times, much had to be done so that students can acquire the requisite knowledge, develop skills, and build on what they have learned. How can we ensure that the learning process is as smooth as possible for everyone involved? How can we evaluate knowledge and skills learned at a distance, and their relevance? Four groups of electronic and electrical engineering students in France were monitored during the containment period in order to provide answers to these questions. Lectures, tutorials, practical work, and projects were carried out using the Microsoft Teams and Zoom video conferencing and chat tools to complement activities made available through the digital work environment. In order to ensure equity among all students, especially in view of the digital divide, open access tools/software/applications have been promoted. In the various surveys completed, the engineering students asserted their complete satisfaction with the learning process, the use of distance tools, and the level of mastery of these tools by their teachers. The results of the various knowledge tests show that, for the same course, distance learning does not reduce the performance of the engineering students. Indeed, they obtained local grades similar to those expected in face-to-face teaching. The results presented in this article are not intended to highlight the virtues of distance education, but rather to open up a debate and reflect more widely on the sustainability of this transformation of education in universities.

Cooperative machine learning has many applications, such as data annotation, where an initial model trained with partially labeled data is used to predict labels for unseen data continuously. Predicted labels with a low confidence value are manually revised to allow the model to be retrained with the predicted and revised data. In this paper, we propose an alternative to this approach: an initial training process called Deep Unsupervised Active Learning. Using the proposed training scheme, a classification model can incrementally acquire new knowledge during the testing phase without manual guidance or correction of decision making. The training process consists of two stages: the first stage of supervised training using a classification model, and an unsupervised active learning stage during the test phase. The labels predicted during the test phase, with high confidence, are continuously used to extend the knowledge base of the model. To optimize the proposed method, the model must have a high initial recognition rate. To this end, we exploited the Visual Geometric Group (VGG16) pre-trained model applied to three datasets: Mathematical Image Analysis (AMI), University of Science and Technology Beijing (USTB2), and Annotated Web Ears (AWE). This approach achieved impressive performance that shows a significant improvement in the recognition rate of the USTB2 dataset by coloring its images using a Generative Adversarial Network (GAN). The obtained performances are interesting compared to the current methods: the recognition rates are 100.00%, 98.33%, and 51.25% for the USTB2, AMI, and AWE datasets, respectively.

Since mid-March 2020, due to the COVID-19 pandemic, higher education has been facing a very uncertain situation, despite the hasty implementation of information and communication technologies for distance and online learning. Hybrid learning, i.e., the mixing of distance and face-to-face learning, seems to be the rule in most universities today. In order to build a post-COVID-19 university education, i.e., one that is increasingly digital and sustainable, it is essential to learn from these years of health crisis. In this context, this paper aims to identify and quantify the main factors affecting mechanical engineering student performance in order to build a generalized linear autoregressive (GLAR) model. This model, which is distinguished by its simplicity and ease of implementation, is responsible for predicting student grades in online learning situations in hybrid environments. The thirty or so variables identified by a previously tested model in 2020–2021, in which distance learning was the exclusive mode of learning, were evaluated in blended learning spaces. Given the low predictive power of the original model, about ten new factors, specific to blended learning, were then identified and tested. The refined version of the GLAR model predicts student grades to within ±1 with a success rate of 63.70%, making it 28.08% more accurate than the model originally created in 2020–2021. Special attention was also given to students whose grade predictions were underestimated and who failed. The methodology presented is applicable to all aspects of the academic process, including students, instructors, and decisionmakers.

Single-Sample Face Recognition (SSFR) is a computer vision challenge. In this scenario, there is only one example from each individual on which to train the system, making it difficult to identify persons in unconstrained environments, mainly when dealing with changes in facial expression, posture, lighting, and occlusion. This paper discusses the relevance of an original method for SSFR, called Multi-Block Color-Binarized Statistical Image Features (MB-C-BSIF), which exploits several kinds of features, namely, local, regional, global, and textured-color characteristics. First, the MB-C-BSIF method decomposes a facial image into three channels (e.g., red, green, and blue), then it divides each channel into equal non-overlapping blocks to select the local facial characteristics that are consequently employed in the classification phase. Finally, the identity is determined by calculating the similarities among the characteristic vectors adopting a distance measurement of the K-nearest neighbors (K-NN) classifier. Extensive experiments on several subsets of the unconstrained Alex and Robert (AR) and Labeled Faces in the Wild (LFW) databases show that the MB-C-BSIF achieves superior and competitive results in unconstrained situations when compared to current state-of-the-art methods, especially when dealing with changes in facial expression, lighting, and occlusion. The average classification accuracies are 96.17% and 99% for the AR database with two specific protocols (i.e., Protocols I and II, respectively), and 38.01% for the challenging LFW database. These performances are clearly superior to those obtained by state-of-the-art methods. Furthermore, the proposed method uses algorithms based only on simple and elementary image processing operations that do not imply higher computational costs as in holistic, sparse or deep learning methods, making it ideal for real-time identification.

Since the beginning of the COVID-19 pandemic, students around the world have seen their schooling completely disrupted. Their teachers had to reorganize in a hurry to be able to give their classes synchronously but at a distance. Thanks to strong political incentives, both at the national and university levels, many tools have been massively deployed: digital learning management systems (e.g. Moodle), collaborative digital platforms (e.g. Google Meets, Microsoft Teams and Zoom), social networks (e.g. Facebook and Twitter) and even the telephone. While the COVID-19 pandemic has highlighted the essential role of digital technologies in higher education, major issues arise regarding the quality of distance education, the learning process itself and the assessment of knowledge and skills. In this study, 81 engineering students in France were followed during several periods of containment in order to provide some answers to these questions. The results of the various knowledge tests carried out at a distance show that the students obtained local scores similar to those expected from face-to-face teaching. According to the results of the satisfaction surveys, for 91.4% of students who had sufficient hardware and software resources, the synchronous approach to e-learning presented few barriers. For the 8.6% of students affected by the digital divide, telephone communications and social networks played a major role in the learning process.