Engineering Applications of Artificial Intelligence最新文献

{"title":"Mutual Information Guided Invertible Image Hiding Network","authors":"Kehan Zhang ,&nbsp;Fen Xiao ,&nbsp;Jingwen Cai ,&nbsp;Xieping Gao","doi":"10.1016/j.engappai.2025.112343","DOIUrl":"10.1016/j.engappai.2025.112343","url":null,"abstract":"<div><div>Image hiding techniques are commonly used for secure communication, copyright protection, and visual privacy. Invertible neural network (INN) have emerged as a promising approach for image steganography, enabling the concealment and recovery of secret images through forward and backward mappings within the network. However, existing methods often face limitations in the accuracy of recovered images due to challenges in estimating the lost information during the forward process. To address this issue, we propose a Mutual Information Guided Invertible Image Hiding Network (MIGIIHNet), which leverages mutual information estimation between the lost information and the stego image in the forward process to guide the backward mapping for reconstruction. Specifically, we propose a lightweight INN with a channel attention feature aggregation module (CAFAM), integrating a channel attention mechanism to optimize the multi-scale aggregation of both low-level and high-level features in a single forward pass. Also, an association learning module (ALM) is designed to model the mutual information between the stego image and the lost information during the forward hiding process. Then, the mutual information is utilized to reconstruct the secret image with high accuracy. Extensive experimental results show that MIGIIHNet outperforms existing state-of-the-art methods in terms of invisibility, security, and recovery accuracy, while maintaining low computational complexity.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112343"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222592","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research on the application of attention mechanism based multi-model fusion in food recommendation platforms","authors":"Linchao Zhang ,&nbsp;Lei Hang","doi":"10.1016/j.engappai.2025.112449","DOIUrl":"10.1016/j.engappai.2025.112449","url":null,"abstract":"<div><div>Smartphone-based food ordering has greatly enhanced convenience in daily life, and the rise of recommendation systems has transformed the functionality and user experience of food delivery applications. Innovations in recommendation algorithms and models have significantly improved the efficiency of food, merchant, and advertisement recommendations on food platforms, leading to higher transaction rates and greater user satisfaction. To further enhance recommendation efficiency, this study introduces a novel multi-model fusion recommendation architecture based on the multi-head self-attention mechanism, utilizing a two-tier structure. The first-tier model (the attention-based homogeneous AutoInt model) acts as a teacher to guide the training of the second-tier Transformer model. This hierarchical approach integrates multiple models through knowledge distillation, significantly improving the accuracy of the recommendation system. The complexity and performance of the proposed architecture were analyzed and applied in a production environment. Testing on a private dataset reveals that the proposed multi-model fusion recommendation architecture significantly enhances recommendation performance across various food platform scenarios, achieving an accuracy of 0.7643, recall of 0.8262, and an F1 score of 0.7936. These results surpass the performance of current state-of-the-art models. Therefore, the proposed architecture is not only highly applicable to food recommendation systems but also has broad applicability in other fields such as retail and entertainment.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112449"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222593","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A diffusion model using semantic and sketch information for anomaly detection","authors":"Li Qin ,&nbsp;Zhenyu Yin ,&nbsp;Feiqing Zhang ,&nbsp;Chunhe Song ,&nbsp;Xiaoqiang Shi","doi":"10.1016/j.engappai.2025.112430","DOIUrl":"10.1016/j.engappai.2025.112430","url":null,"abstract":"<div><div>In anomaly detection, methods that employ diffusion models for anomaly localization and reconstruction have demonstrated significant achievements. However, these methods face challenges such as the misclassification of multiple types of anomalies and the inability to effectively reconstruct large-scale anomalies due to the absence of semantic and sketch information from the original images. To tackle these challenges, we propose a framework, A Diffusion Model using Semantic and Sketch Information for Anomaly Detection (DSAD), which includes a semantic and sketch-guided network (SSG), a pre-trained autoencoder, and Stable Diffusion (SD). Initially, within SSG, we introduce a Semantic <span><math><mi>&amp;</mi></math></span> Sketch Feature Fusion Module to enhance the model’s comprehension of the original images and present a Multi-scale Feature Fusion Module to maximize reconstruction accuracy. Subsequently, we connect SSG with the denoising network in SD in order to guide the network in reconstructing anomalous regions. Experiments on MVTec-AD dataset demonstrate the effectiveness of our approach which surpasses the state-of-the-art methods. The dataset and code are available at <span><span>https://github.com/QinLi-STUDY/DSAD/tree/master</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112430"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159651","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A transfer learning method of collaborating random walk and adaptive instance normalization for inscription image denoising","authors":"Erhu Zhang ,&nbsp;Yunjing Liu ,&nbsp;Guangfeng Lin ,&nbsp;Jinghong Duan","doi":"10.1016/j.engappai.2025.112458","DOIUrl":"10.1016/j.engappai.2025.112458","url":null,"abstract":"<div><div>Mess noise hinders reading and understanding of inscriptions in images. For image restoration from noise-corrupted images, existing network-learning-based methods can construct an excellent model to generate noise patterns. However, the performance of such models is degraded owing to the lack of high-quality training data and the complex noise pattern in inscription images, e.g., mixed noise with multiple levels. Herein, we first propose a novel noise generation model that can produce more realistic synthetic noise images using the random walk algorithm. Then, we propose an explainable inscription image denoising network using a variational inference model, where the joint distribution of clean-noise image pairs is approximated in a dual adversarial manner. The proposed network exhibits improved generalizability and adaptability to different noise characteristics using an estimated noise map and adaptive instance normalization. Finally, we introduce a transfer learning scheme to migrate the network learned from the synthetic noise image domain to a real-inscription image domain with a limited number of real-inscription images. The proposed method outperforms state-of-the-art methods.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112458"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159777","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detecting worker loss of balance events from point cloud sequence using unsupervised motion-pose learning","authors":"Mingyu Zhang,&nbsp;Lei Wang,&nbsp;Yinong Hu,&nbsp;Shuai Han,&nbsp;Jiawen Zhang,&nbsp;Heng Li","doi":"10.1016/j.engappai.2025.112512","DOIUrl":"10.1016/j.engappai.2025.112512","url":null,"abstract":"<div><div>Workers' loss of balance (LB), such as slip and trip, may lead to severe injuries and even fatalities. Existing methods for detecting LB typically rely on wearable sensors and focus on specific body parts. This study introduces a novel, non-contact approach utilizing light detection and ranging (LiDAR) technology to detect LB events. By capturing full-body point cloud data, the proposed method extracts both static pose and dynamic motion features across multiple body sections and detects LB events through unsupervised learning. The high-dimensional point cloud sequence is transformed into interpretable gait features, enabling effective unsupervised learning through sequence reconstruction. A two-stream network and fusion strategy are also developed to combine pose and motion features for final LB detection. Experiments with various LB events demonstrate the method's effectiveness, achieving an F1 score of 0.98 and a recall of 0.98. Our analysis reveals that integrating features from multiple body parts and the fusion of pose and motion information significantly enhances detection performance. This study offers a promising alternative to traditional methods, providing effective, non-intrusive monitoring of worker safety in dynamic construction environments.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112512"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145160240","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Sparse and robust elastic net support vector machine with bounded concave loss for large-scale problems","authors":"Huajun Wang,&nbsp;Wenqian Li","doi":"10.1016/j.engappai.2025.112352","DOIUrl":"10.1016/j.engappai.2025.112352","url":null,"abstract":"<div><div>The elastic net support vector machine is an extensively employed method for addressing a range of classification tasks. Nevertheless, a significant drawback of the elastic net support vector machine is its high computational cost when dealing with large-scale classification problems. To address this drawback, we first introduce an innovative non-convex elastic net support vector machine model that employs our newly created bounded concave loss function, which effectively attains both sparsity and robustness. Based on proximal stationary point, we have effectively constructed an innovative optimality theory tailored for our newly created elastic net support vector machine model. By leveraging the innovative optimality theory, we have successfully developed a new and exceptionally effective algorithm designed to enhance computational efficiency through the division of the entire dataset into two distinct categories: working sets and non-working sets. During each learning cycle, the parameters associated with the non-working set remain unchanged. In contrast, the parameters related to the working set are subject to updates. Consequently, our new algorithm facilitates quicker modifications on smaller datasets, improving runtime efficiency and lowering computational complexity. Numerical experiments have demonstrated significant efficiency, particularly regarding computational speed, the number of support vectors, and classification accuracy, surpassing eleven other leading solvers.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112352"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145222590","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Detecting code paraphrased by large language models using coding style features","authors":"Shinwoo Park ,&nbsp;Hyundong Jin ,&nbsp;Jeong-won Cha ,&nbsp;Yo-Sub Han","doi":"10.1016/j.engappai.2025.112454","DOIUrl":"10.1016/j.engappai.2025.112454","url":null,"abstract":"<div><div>Recent progress in large language models (LLMs) for code generation has raised serious concerns about intellectual property protection. Malicious users can exploit LLMs to produce paraphrased versions of proprietary code that closely resemble the original. While the potential for LLM-assisted code paraphrasing continues to grow, research on detecting it remains limited, underscoring an urgent need for a detection system. We respond to this need by proposing two tasks. The first task is to detect whether code generated by an LLM is a paraphrased version of original human-written code. The second task is to identify which LLM is used to paraphrase the original code. For these tasks, we construct a dataset consisting of pairs of human-written code and LLM-paraphrased code using various LLMs.</div><div>We statistically confirm significant differences in the coding styles of human-written and LLM-paraphrased code, particularly in terms of naming consistency, code structure, and readability. Based on these findings, we develop a detection method that identifies paraphrase relationships between human-written and LLM-generated code, and discover which LLM is used for the paraphrasing. Our detection method outperforms the best baselines in two tasks, improving F1 scores by 2.64% and 15.17% while achieving speedups of 1,343x and 213x, respectively.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112454"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145159633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cooperative neural networks for inverse design: Integrating denoising autoencoder and surrogate model for partial design variable imputation","authors":"Agung Nugraha ,&nbsp;Hyerin Kwon ,&nbsp;Gyeongho Park ,&nbsp;Jihwan Lee","doi":"10.1016/j.engappai.2025.112453","DOIUrl":"10.1016/j.engappai.2025.112453","url":null,"abstract":"<div><div>Data-driven inverse design is an engineering approach where target performance criteria are specified upfront, leading to the derivation of design solutions that meet these criteria. While recent research focuses on generating complete design solutions using generative models, these approaches struggle with partial design variables and constraints that predetermine certain variables. Additionally, generative models are data-intensive and prone to overfitting with limited datasets. To address these limitations, this paper proposes a Cooperative Neural Network architecture comprising two key components: the Imputation Model and the Surrogate Model. These components collaborate to optimize design solutions while adhering to predefined performance criteria. The framework’s effectiveness is demonstrated through a case study on Glass Run Channel (GRC) designs from a Korean automotive manufacturer. Results show the architecture proficiently imputes undetermined variables and ensures the designs meet desired performance metrics, achieving Mean Squared Error (MSE) reductions of up to 98 % and R-squared values of 0.997–0.999 in initial tests. It remains robust in diverse scenarios, achieving up to 95.65 % MSE reduction and R-squared values of 0.995–0.999 for cases with the most undetermined variables, and up to 94.68 % MSE reduction with R-squared values of 0.983–0.995 for the smallest training datasets. This framework reduces design cycle times and enhances engineering design efficiency, offering a robust solution to limitations in traditional methods reliant on physical prototyping and iterative testing.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112453"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145160237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unlocking few-shot wind speed prediction through a novel end-to-end transfer learning paradigm based on decomposition and gating information fusion","authors":"Xiaoyue Dong ,&nbsp;Zhirui Tian","doi":"10.1016/j.engappai.2025.112435","DOIUrl":"10.1016/j.engappai.2025.112435","url":null,"abstract":"<div><div>Accurate wind speed prediction is one of the key technologies for achieving intelligent and sustainable development in the engineering field. In the field of wind speed prediction, we are confronted with a challenging few-shot prediction problem. Specifically, due to the fact that some wind turbines in wind farms are newly established or there is data loss during the data collection process, these turbines only contain a small amount of wind speed data. This scarcity of data poses great difficulties for the prediction work, and traditional prediction methods often fail to achieve the desired prediction accuracy. In order to overcome the above difficulties, we propose an novel prediction paradigm of end-to-end transfer learning based on data decomposition and gated information fusion. We use the Fourier transform to find the source domain similar to the target domain to achieve feature alignment. Then, we pre-train the model on the source domain and transfer this model to the target domain, thus solving the problem of low prediction accuracy when directly predicting the target domain. In the first step, the data is decomposed and denoised by using the Variational Mode Decomposition. According to the sample entropy, the decomposed data is reorganized into three frequency components. Each component is input as an independent channel into the end-to-end prediction model. Firstly, the features of each channel are expanded to a high-dimensional space through the Multilayer Perceptron. Then, the gating mechanism is utilized to mix the features of the three channels into the features of one channel, thus achieving information fusion. Finally, the prediction result of the end-to-end model is output through the Gated Recurrent Unit. In the second step, the model pre-trained on the source domain is transferred to the small-sample target domain. The Dynamic Time Warping and cosine similarity are used to quantify the similarity of each channel between the two domains. The parameters of the channels with high similarity are locked, and at the same time, the parameters of other channels are fine-tuned to output the final prediction result. In addition, multiple sets of comparative experiments conducted using the wind speed data from wind farms in Queensland, Australia, have demonstrated the superiority of this prediction paradigm. Our strategy outperforms various baseline models in all three sets of data. Moreover, ablation experiments have proven the effectiveness of each component in this framework in improving prediction accuracy, opening up a new path for solving the difficult problem of few-shot prediction in practical engineering.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112435"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145160235","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Transformer architecture with illumination aware mechanisms for low-light image enhancement via Retinex decomposition","authors":"Zixuan Wang ,&nbsp;Gang Liu ,&nbsp;Hanlin Xu ,&nbsp;Yao Qian ,&nbsp;Rui Chang ,&nbsp;Durga Prasad Bavirisetti","doi":"10.1016/j.engappai.2025.112414","DOIUrl":"10.1016/j.engappai.2025.112414","url":null,"abstract":"<div><div>Enhancing low-light images is a complex task that involves not only restoring brightness but also preserving color fidelity and reducing noise interference. In this paper, we propose a novel Retinex-based Transformer Model with Illumination Aware Mechanisms (TIMRetinex-Net), which achieves physically interpretable modeling through a decomposition network guided by Retinex theory. To adapt to light variations in different regions, we randomly apply gamma transformations to several subregions of the illumination component and use a Color Estimation Module to capture the color global distribution of the natural scene in the reflection component. By modeling the color global distribution and repairing the degraded regions collaboratively, we alleviate the issue of being highly sensitive to data usage during training and improve the model’s ability to handle unknown scenes. The Illumination and Reflection Adjustment Transformer Network (IRAT-Net) produces enhanced images, achieving a balanced enhancement of detail and color. In addition, IRAT-Net incorporates an attention mechanism into the feature extraction layer and introduces the Illumination-Guided Information Aggregation Module to adaptively estimate lighting conditions. In the field of image processing, our method based on artificial intelligence was evaluated on five datasets and compared with twelve state-of-the-art methods. The results demonstrated strong alignment with the ground truth, with our method achieving superior performance in both subjective and objective assessments.</div></div>","PeriodicalId":50523,"journal":{"name":"Engineering Applications of Artificial Intelligence","volume":"162 ","pages":"Article 112414"},"PeriodicalIF":8.0,"publicationDate":"2025-09-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145160236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}