Explore the vast range of titles available.

With the development of smart agriculture, deep learning is playing an increasingly important role in crop disease recognition. The existing crop disease recognition models are mainly based on convolutional neural networks (CNN). Although traditional CNN models have excellent performance in modeling local relationships, it is difficult to extract global features. This study combines the advantages of CNN in extracting local disease information and vision transformer in obtaining global receptive fields to design a hybrid model called MSCVT. The model incorporates the multiscale self-attention module, which combines multiscale convolution and self-attention mechanisms and enables the fusion of local and global features at both the shallow and deep levels of the model. In addition, the model uses the inverted residual block to replace normal convolution to maintain a low number of parameters. To verify the validity and adaptability of MSCVT in the crop disease dataset, experiments were conducted in the PlantVillage dataset and the Apple Leaf Pathology dataset, and obtained results with recognition accuracies of 99.86% and 97.50%, respectively. In comparison with other CNN models, the proposed model achieved advanced performance in both cases. The experimental results show that MSCVT can obtain high recognition accuracy in crop disease recognition and shows excellent adaptability in multidisease recognition and small-scale disease recognition.

Fault diagnosis is crucial for repairing aircraft and ensuring their proper functioning. However, with the higher complexity of aircraft, some traditional diagnosis methods that rely on experience are becoming less effective. Therefore, this paper explores the construction and application of an aircraft fault knowledge graph to improve the efficiency of fault diagnosis for maintenance engineers. Firstly, this paper analyzes the knowledge elements required for aircraft fault diagnosis, and defines a schema layer of a fault knowledge graph. Secondly, with deep learning as the main method and heuristic rules as the auxiliary method, fault knowledge is extracted from structured and unstructured fault data, and a fault knowledge graph for a certain type of craft is constructed. Finally, a fault question-answering system based on a fault knowledge graph was developed, which can accurately answer questions from maintenance engineers. The practical implementation of our proposed methodology highlights how knowledge graphs provide an effective means of managing aircraft fault knowledge, ultimately assisting engineers in identifying fault roots accurately and quickly.

Violence is a serious threat to societal health; preventing violence in airports, airplanes, and spacecraft is crucial. This study proposes the Violence-YOLO model to detect violence accurately in real time in complex environments, enhancing public safety. The model is based on YOLOv9’s Generalized Efficient Layer Aggregation Network (GELAN-C). A multilayer SimAM is incorporated into GELAN’s neck to identify attention regions in the scene. YOLOv9 modules are combined with RepGhostNet and GhostNet. Two modules, RepNCSPELAN4_GB and RepNCSPELAN4_RGB, are innovatively proposed and introduced. The shallow convolution in the backbone is replaced with GhostConv, reducing computational complexity. Additionally, an ultra-lightweight upsampler, Dysample, is introduced to enhance performance and reduce overhead. Finally, Focaler-IoU addresses the neglect of simple and difficult samples, improving training accuracy. The datasets are derived from RWF-2000 and Hockey. Experimental results show that Violence-YOLO outperforms GELAN-C. mAP@0.5 increases by 0.9%, computational load decreases by 12.3%, and model size is reduced by 12.4%, which is significant for embedded hardware such as the Raspberry Pi. Violence-YOLO can be deployed to monitor public places such as airports, effectively handling complex backgrounds and ensuring accurate and fast detection of violent behavior. In addition, we achieved 84.4% mAP on the Pascal VOC dataset, which is a significant reduction in model parameters compared to the previously refined detector. This study offers insights for real-time detection of violent behaviors in public environments.

This paper presents an empirical study on how knowledge management practices and innovation sources affect product innovation performance, among the 152 manufacturers in the low- and high- tech industries in China. The results indicate that external innovation sources are positively correlated with innovation activities and new product performance. Intellectual Property (IP) and knowledge management practices (KMP) are positively correlated with innovation activities, and KMP is positively correlated with innovation sources. The dual effect of KMP shows its indispensable effect on the new product development for both high-tech and low-tech firms, but for low-tech firms, such effect is relatively weak. This empirical study shows that IP management is critical to high-tech but not low-tech firms. We also found that, for innovation activities, low-tech depends on the external sources of innovation whilst high-tech firms do not.

To address critical challenges in X-ray contraband detection—including severe class imbalance in existing datasets, scarcity of high-quality annotated data, and poor model adaptability to complex scenarios—this study first constructs a balanced X-ray contraband detection dataset. Derived from the SIXray and PIDray datasets, the balanced dataset comprises 13,728 images covering 12 different contraband categories. To resolve class imbalance, a Class-Specific Augmentation Framework (CSAF) with four physical transformations and random undersampling are adopted, ensuring approximately 1,500 samples per category for uniform class distribution. Two improved models (ASEA-Net and CSEC-Net) based on YOLOv11s are proposed for lightweight and high-precision contraband detection tasks. Experiments on the balanced dataset show that ASEA-Net achieves 95.78% accuracy and 93.55% mAP@50, outperforming YOLOv11s by 1.46% and 1.37% respectively with 13.37% fewer parameters; CSEC-Net reduces parameters by 39.91% and FLOPs by 40.38% compared to YOLOv11s, enabling deployment on resource-constrained edge devices. Both models exhibit strong performance in complex scenarios, validating the value of the balanced dataset and the effectiveness of the proposed models for X-ray contraband detection.

In the world of social networking, consumers tend to refer to expert comments or product reviews before making buying decisions. There is much useful information available on many social networking sites for consumers to make product comparisons. Sentiment analysis is considered appropriate for summarising the opinions. However, the sentences posted online are generally short, which sometimes contains both positive and negative word in the same post. Thus, it may not be sufficient to determine the sentiment polarity of a post by merely counting the number of sentiment words, summing up or averaging the associated scores of sentiment words. In this paper, an unsupervised learning technique, k-means, in conjunction with sentiment analysis, is proposed for assessing public opinions. The proposed approach offers the product designers a tool to promptly determine the critical design criteria for new product planning in the process of new product development by evaluating the user-generated content. The case implementation proves the applicability of the proposed approach.

With the development of the Internet, people are surrounded by various types of information daily, including obscene videos. The quantity of such videos is increasing daily, making the detection and filtering of this information a crucial step in preventing its spread. However, a significant challenge remains in detecting obscene information in obscure scenarios, like indecent behavior occurring while wearing normal clothing, causing significant negative impacts, such as harmful influence on children. To address this issue, an innovative multi frame obscene video detection base on ViT is proposed by this manuscript per the authors, aiming to automatically detect and filter obscene content in videos. Extensive experiments conducted on the public NPDI dataset demonstrate that this method achieves better results than existing state-of-the-art methods, achieving 96.2%. Additionally, it achieves satisfactory classification accuracy on a dataset of obscure obscene videos.This provides a powerful tool for future video censorship and protects minors and the general public.

Identifying drugs from surveillance or other videos presents challenges such as small target sizes, class imbalance, and similarities to other objects. Additionally, the hardware used to capture videos and the video resolution and clarity limit model scalability, leading to poor detection accuracy in traditional models. To address this issue, we propose an improved YOLOv8s-based model. The experimental outcomes reveal that the improved YOLOv8s model attains a precision of 95.1% and a mAP@50 of 87.4% in drug detection and identification, representing improvements of 3.0% and 2.2% over the original YOLOv8s model. The proposed improvements to YOLOv8s effectively boost detection accuracy and recognition rates while preserving high efficiency. This model demonstrates superior overall detection performance compared to other algorithms, providing fresh perspectives and methods for advancing research and applications in drug detection and recognition.

The rapid development of the Internet has facilitated expression, sharing, and interaction on social networks, but some speech may contain harmful discrimination. Therefore, it is crucial to classify such speech. In this paper, we collected discriminatory data from Sina Weibo and propose the improved Synthetic Minority Over-sampling Technique (SMOTE) algorithm based on Latent Dirichlet Allocation (LDA) to improve data quality and balance. And we propose a new integration method integrating Support Vector Machine (SVM) and Random Forest (RF). The experimental results demonstrate that the integrated model exhibits enhanced precision, recall, and F1 score by 6.0%, 5.4%, and 5.7%, respectively, in comparison with SVM alone. Moreover, it exhibits the best performance in comparison with other machine learning methods. Furthermore, the positive impact of improved SMOTE and this integrated method on model classification is also confirmed in ablation experiments.

COVID-19 infection has been hypothesized to precipitate venous and arterial clotting events more frequently than other illnesses. We demonstrate this increased risk of blood clots by comparing rates of venous and arterial clotting events in 4400 hospitalized COVID-19 patients in a large multisite clinical network in the United States examined from April through June of 2020, to patients hospitalized for non-COVID illness and influenza during the same time period and in 2019. We demonstrate that COVID-19 increases the risk of venous thrombosis by two-fold compared to the general inpatient population and compared to people with influenza infection. Arterial and venous thrombosis were both common occurrences among patients with COVID-19 infection. Risk factors for thrombosis included male gender, older age, and diabetes. Patients with venous or arterial thrombosis had high rates of admission to the ICU, re-admission to the hospital, and death. Given the ongoing scientific discussion about the impact of clotting on COVID-19 disease progression, these results highlight the need to further elucidate the role of anticoagulation in COVID-19 patients, particularly outside the intensive care unit setting. Additionally, concerns regarding clotting and COVID-19 vaccines highlight the importance of addressing the alarmingly high rate of clotting events during actual COVID-19 infection when weighing the risks and benefits of vaccination.