The existing deep learning-based Personal Protective Equipment (PPE) detectors can only detect limited types of PPE and their performance needs to be improved, particularly for their deployment on real construction sites. This paper introduces an approach to train and evaluate eight deep learning detectors, for real application purposes, based on You Only Look Once (YOLO) architectures for six classes, including helmets with four colours, person, and vest. Meanwhile, a dedicated high-quality dataset, CHV, consisting of 1330 images, is constructed by considering real construction site background, different gestures, varied angles and distances, and multi PPE classes. The comparison result among the eight models shows that YOLO v5x has the best mAP (86.55%), and YOLO v5s has the fastest speed (52 FPS) on GPU. The detection accuracy of helmet classes on blurred faces decreases by 7%, while there is no effect on other person and vest classes. And the proposed detectors trained on the CHV dataset have a superior performance compared to other deep learning approaches on the same datasets. The novel multiclass CHV dataset is open for public use.

Carbon emissions present a pressing challenge to the traditional construction industry, urging a fundamental shift towards more sustainable practices and materials. Recent advances in sensors, data fusion techniques, and artificial intelligence have enabled integrated digital technologies (e.g., digital twins) as a promising trend to achieve emission reduction and net-zero. While digital twins in the construction sector have shown rapid growth in recent years, most applications focus on the improvement of productivity, safety and management. There is a lack of critical review and discussion of state-of-the-art digital twins to improve sustainability in this sector, particularly in reducing carbon emissions. This paper reviews the existing research where digital twins have been directly used to enhance sustainability throughout the entire life cycle of a building (including design, construction, operation and maintenance, renovation, and demolition). Additionally, we introduce a conceptual framework for this industry, which involves the elements of the entire digital twin implementation process, and discuss the challenges faced during deployment, along with potential research opportunities. A proof-of-concept example is also presented to demonstrate the validity of the proposed conceptual framework and potential of digital twins for enhanced sustainability. This study aims to inspire more forward-thinking research and innovation to fully exploit digital twin technologies and transform the traditional construction industry into a more sustainable sector.

Characterisation of the driver's non-driving activities (NDAs) is of great importance to the design of the take-over control strategy in Level 3 automation. Gaze estimation is a typical approach to monitor the driver's behaviour since the eye gaze is normally engaged with the human activities. However, current eye gaze tracking techniques are either costly or intrusive which limits their applicability in vehicles. This paper proposes a low-cost and non-intrusive dual-cameras based gaze mapping system that visualises the driver's gaze using a heat map. The challenges introduced by complex head movement during NDAs and camera distortion are addressed by proposing a nonlinear polynomial model to establish the relationship between the face features and eye gaze on the simulated driver's view. The Root Mean Square Error of this system in the in-vehicle experiment for the X and Y direction is 7.80±5.99 pixel and 4.64±3.47 pixel respectively with the image resolution of 1440 x 1080 pixels. This system is successfully demonstrated to evaluate three NDAs with visual attention. This technique, acting as a generic tool to monitor driver's visual attention, will have wide applications on NDA characterisation for intelligent design of take over strategy and driving environment awareness for current and future automated vehicles.

The rapid development of large language models (LLMs) and multimodal large models (MLMs) has introduced transformative opportunities for autonomous driving systems. These advanced models provide robust support for the realization of more intelligent, safer, and efficient autonomous driving. In this paper, we present a systematic review on the integration of LLMs and MLMs in autonomous driving systems. First, we provide an overview of the evolution of LLMs and MLMs, along with a detailed analysis of the architecture of autonomous driving systems. Next, we explore the applications of LLMs and MLMs in key components such as perception, prediction, decision making, planning, multitask processing, and human–machine interaction. Additionally, this paper reviews the core technologies involved in integrating LLMs and MLMs with autonomous driving systems, including multimodal fusion, knowledge distillation, prompt engineering, and supervised fine tuning. Finally, we provide an in-depth analysis of the major challenges faced by autonomous driving systems powered by large models, offering new perspectives for future research. Compared to existing review articles, this paper not only systematically examines the specific applications of LLMs and MLMs in autonomous driving systems but also delves into the key technologies and potential challenges involved in their integration. By comprehensively organizing and analyzing the current literature, this review highlights the application potential of large models in autonomous driving and offers insights and recommendations for improving system safety and efficiency.

Alzheimer's disease (AD) and Parkinson's disease (PD) are two of the most common forms of neurodegenerative diseases. The literature suggests that effective brain connectivity (EBC) has the potential to track differences between AD, PD and healthy controls (HC). However, how to effectively use EBC estimations for the research of disease diagnosis remains an open problem. To deal with complex brain networks, graph neural network (GNN) has been increasingly popular in very recent years and the effectiveness of combining EBC and GNN techniques has been unexplored in the field of dementia diagnosis. In this study, a novel directed structure learning GNN (DSL-GNN) was developed and performed on the imaging of EBC estimations and power spectrum density (PSD) features. In comparison to the previous studies on GNN, our proposed approach enhanced the functionality for processing directional information, which builds the basis for more efficiently performing GNN on EBC. Another contribution of this study is the creation of a new framework for applying univariate and multivariate features simultaneously in a classification task. The proposed framework and DSL-GNN are validated in four discrimination tasks and our approach exhibited the best performance, against the existing methods, with the highest accuracy of 94.0% (AD vs. HC), 94.2% (PD vs. HC), 97.4% (AD vs. PD) and 93.0% (AD vs. PD vs. HC). In a word, this research provides a robust analytical framework to deal with complex brain networks containing causal directional information and implies promising potential in the diagnosis of two of the most common neurodegenerative conditions.

Quantitative defect and damage reconstruction play a critical role in industrial quality management. Accurate defect characterisation in Infrared Thermography (IRT), as one of the widely used Non-Destructive Testing (NDT) techniques, always demands adequate pre-knowledge which poses a challenge to automatic decision-making in maintenance. This paper presents an automatic and accurate defect profile reconstruction method, taking advantage of deep learning Neural Networks (NN). Initially, a fast Finite Element Modelling (FEM) simulation of IRT is introduced for defective specimen simulation. Mask Region-based Convolution NN (Mask-RCNN) is proposed to detect and segment the defect using a single thermal frame. A dataset with a single-type-shape defect is tested to validate the feasibility. Then, a dataset with three mixed shapes of defect is inspected to evaluate the method’s capability on the defect profile reconstruction, where an accuracy over 90% on Intersection over Union (IoU) is achieved. The results are compared with several state-of-the-art of post-processing methods in IRT to demonstrate the superiority at detailed defect corners and edges. This research lays solid evidence that AI deep learning algorithms can be utilised to provide accurate defect profile reconstruction in thermography NDT, which will contribute to the research community in material degradation analysis and structural health monitoring.