Photovoltaic (PV) industries are susceptible to manufacturing defects within their solar cells. To accurately evaluate the efficacy of solar PV modules, the identification of manufacturing defects is imperative. Conventional industrial defect inspections predominantly rely on highly skilled inspectors conducting manual defect assessments, leading to sporadic and subjective identification outcomes. Deep-learning-based fault detection in PV or solar cells has emerged as a primary research area due to its superior efficiency and applicability. Hence, this study introduces a SegFormer-based fault detection framework to automate the visual defect inspection process in PV modules, complete with defect pseudocolorization. The proposed SegFormer-based framework effectively classifies defects into five categories: crack defects, front grid defects, interconnect defects, contact corrosion defects, and bright disconnect. Moreover, a comparative analysis is performed between the SegFormer model and the state-of-the-art fault detection algorithms, such as Deeplab v3, UNET, Deeplab v3+, PAN, PSPNet, and feature pyramid network (FPN). The experimental results reveal that the proposed SegFormer-based framework achieves highly encouraging performance, with a pixelwise accuracy of 96.24%, a weighted F1-score of 96.22%, an unweighted F1-score of 81.96%, and a mean intersection over union of 56.54%, outperforming other existing methods.

Several efforts have been made to develop effective and robust vision-based solutions for human action recognition in aerial videos. Generally, the existing methods rely on the extraction of either spatial features (patch-based methods) or skeletal key points (pose-based methods) that are fed to a classifier. Unlike the patch-based methods, the pose-based methods are generally regarded to be more robust to background changes and computationally efficient. Moreover, at the classification stage, the use of deep networks has generated significant interest within the community; however, the need remains to develop accurate and computationally effective deep learning-based solutions. To this end, this paper proposes a lightweight Transformer network-based method for human action recognition in aerial videos using the skeletal keypoints extracted using YOLOv8. The effectiveness of the proposed method is shown on a well-known public dataset containing 13 action classes, achieving very encouraging performance in terms of accuracy and computational cost as compared to several existing related methods.

The accurate detection and recognition of human actions play a pivotal role in aerial surveillance, enabling the identification of potential threats and suspicious behavior. Several approaches have been presented to address this problem, but the limitation still remains in devising an accurate and robust solution. To this end, this paper presents an effective action recognition framework for aerial surveillance, employing the YOLOv8-Pose keypoints extraction algorithm and a customized sequential ConvLSTM (Convolutional Long Short-Term Memory) model for classifying the action. We performed a detailed experimental evaluation and comparison on the publicly available Drone Action dataset. The evaluation and comparison of the proposed framework with several existing approaches on the publicly available Drone Action dataset demonstrate its effectiveness, achieving a very encouraging performance. The overall accuracy of the framework on three provided dataset splits is 74%, 80%, and 70%, with a mean accuracy of 74.67%. Indeed, the proposed system effectively captures the spatial and temporal dynamics of human actions, providing a robust solution for aerial action recognition.

The integration of Internet of Things (IoT) technology and medical devices in healthcare is termed the Internet of Medical Things (IoMT). This advancement holds promise for numerous applications aimed at mitigating the risk of loss of life through physiological signal monitoring. As the number of road accidents is rapidly increasing, a substantial number of car crashes occur due to medical conditions. Therefore, the need remains to develop an effective solution to enable the prevention of such accidents for enhanced road safety. Unlike existing approaches, this paper proposes a holistic IoMT‐based non‐invasive driver health monitoring system (DHMS) to monitor important vital signs for detecting abnormal health conditions. The proposed system consists of an embedded system, edge computing, cloud computing, and a mobile application with an alert system, to offer an end‐to‐end unified solution for driver physiological signal monitoring to detect abnormal health conditions that might lead to a road accident. The system is particularly suited to aid (elderly) people with medical conditions and can also be used for public transport to ensure passenger safety. A detailed experimental evaluation of the proposed system has been performed and its performance accuracy compared with standard medical devices, along with quality factors including usability, portability, and effective sensor placement.

In brain-computer interfacing, the SSVEP (steady-state visual evoked potential) method serves to foster collaboration between humans and robots. SSVEP-based detection methods require complex multichannel data acquisition, making them difficult to deploy due to discomfort during extended use and the complexity of the algorithms involved. On the other hand, single-channel setup offers simplicity and ease of use. However, in a single channel, achieving encouraging performance in the SD (subject-dependent) scenario is challenging, and accuracy drops further in the SI (subject-independent) scenario. This requires the development of a generalized approach to improve performance in both scenarios. This study proposes (VMD-DNN) to detect SSVEP in single-channel setups for SD and SI scenarios. The novelty of the proposed method lies in utilizing VMD (Variational Mode Decomposition) as a preprocessor, leveraging harmonic information and Kurtosis of the cross-correlation function to select harmonics from VMD decomposed signal. The preprocessed reconstructed signal uses complex spectrum features as input to the DNN for classification. The results show an average accuracy of 93%, 95.3% in SD and 79%, 92.33% in SI scenarios tested on two publicly available datasets, respectively. The ITR (Information transfer rate) was 67.50 bit/min, 92.31 bit/min for SD, and 46.13 bit/min, 85.94 bit/min for SI for both datasets, respectively. In SD, accuracy is improved by 3.34% and 5%, and ITR by 8.87% and 12.91% over baseline methods for both datasets respectively. The proposed VMD-DNN model is effective, with improved performance and lower computational complexity. The robust single-channel approach makes it user-friendly for human-robot collaboration.