Search Results (2,960)

The underwater imaging process is often hindered by high noise levels, blurring, and color distortion due to light scattering, absorption, and suspended particles in the water. To address the challenges of image enhancement in complex underwater environments, this paper proposes an underwater image color correction and detail enhancement model based on an improved Cycle-consistent Generative Adversarial Network (CycleGAN), named LPIPS-MAFA CycleGAN (LM-CycleGAN). The model integrates a Multi-scale Adaptive Fusion Attention (MAFA) mechanism into the generator architecture to enhance its ability to perceive image details. At the same time, the Learned Perceptual Image Patch Similarity (LPIPS) is introduced into the loss function to make the training process more focused on the structural information of the image. Experiments conducted on the public datasets UIEB and EUVP demonstrate that LM-CycleGAN achieves significant improvements in Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), Average Gradient (AG), Underwater Color Image Quality Evaluation (UCIQE), and Underwater Image Quality Measure (UIQM). Moreover, the model excels in color correction and fidelity, successfully avoiding issues such as red checkerboard artifacts and blurred edge details commonly observed in reconstructed images generated by traditional CycleGAN approaches. Full article

To overcome the problems in existing infrared remote sensing image generation methods, which make it difficult to combine high fidelity and high efficiency, we propose a High-Fidelity Infrared Remote Sensing Image Generation Method Coupled with the Global Radiation Scattering Mechanism and Pix2PixGAN (HFIRSIGM_GRSMP) in this paper. Firstly, based on the global radiation scattering mechanism, the HFIRSIGM_GRSMP model is constructed to address the problem of accurately characterizing factors that affect fidelity—such as the random distribution of the radiation field, multipath scattering, and nonlinear changes—through the innovative fusion of physical models and deep learning. This model accurately characterizes the complex radiation field distribution and the image detail-feature mapping relationship from visible-to-infrared remote sensing. Then, 8000 pairs of image datasets were constructed based on Landsat 8 and Sentinel-2 satellite data. Finally, the experiment demonstrates that the average SSIM of images generated using HFIRSIGM_GRSMP reaches 89.16%, and all evaluation metrics show significant improvement compared to the contrast models. More importantly, this method demonstrates high accuracy and strong adaptability in generating short-wave, mid-wave, and long-wave infrared remote sensing images. This method provides a more comprehensive solution for generating high-fidelity infrared remote sensing images. Full article

As the complexity of GIS data continues to increase, there is a growing demand for automated map generalization. As end-to-end generative models, GAN models offer new solutions for automated map generalization. This study explores the impact of different map symbolization configurations on generative models, specifically using CycleGAN for line feature generalization. The quality of the generated results was assessed by constructing various symbolization datasets (line width, type, and color) and evaluating CycleGAN’s performance using metrics such as the MSE, SSIM, and PSNR. The results indicate that moderate line widths (0.5–1) yield better detail preservation, and different line types (framed lines and dashed lines) can highlight feature boundaries and enhance visual perception. By contrast, high-contrast color schemes enhance feature differentiation but increase pixel-level errors. This study concludes that generative models can maintain the geometric structure and spatial distribution of line features, but it is crucial to choose more suitable line features for different scenarios to meet detail requirements, ensuring high-quality outputs under diverse configurations. Full article

Background: Lung cancer remains a significant health concern, and the effectiveness of early detection significantly enhances patient survival rates. Identifying lung tumors with high precision is a challenge due to the complex nature of tumor structures and the surrounding lung tissues. Methods: To address these hurdles, this paper presents an innovative three-step approach that leverages Generative Adversarial Networks (GAN), Long Short-Term Memory (LSTM), and VGG16 algorithms for the accurate reconstruction of three-dimensional (3D) lung tumor images. The first challenge we address is the accurate segmentation of lung tissues from CT images, a task complicated by the overwhelming presence of non-lung pixels, which can lead to classifier imbalance. Our solution employs a GAN model trained with a reinforcement learning (RL)-based algorithm to mitigate this imbalance and enhance segmentation accuracy. The second challenge involves precisely detecting tumors within the segmented lung regions. We introduce a second GAN model with a novel loss function that significantly improves tumor detection accuracy. Following successful segmentation and tumor detection, the VGG16 algorithm is utilized for feature extraction, preparing the data for the final 3D reconstruction. These features are then processed through an LSTM network and converted into a format suitable for the reconstructive GAN. This GAN, equipped with dilated convolution layers in its discriminator, captures extensive contextual information, enabling the accurate reconstruction of the tumor’s 3D structure. Results: The effectiveness of our method is demonstrated through rigorous evaluation against established techniques using the LIDC-IDRI dataset and standard performance metrics, showcasing its superior performance and potential for enhancing early lung cancer detection. Conclusions:This study highlights the benefits of combining GANs, LSTM, and VGG16 into a unified framework. This approach significantly improves the accuracy of detecting and reconstructing lung tumors, promising to enhance diagnostic methods and patient results in lung cancer treatment. Full article

We propose the ATGT3D an Animatable Texture Generation and Tracking for 3D Avatars, featuring the innovative design of the Eye Diffusion Module (EDM) and Pose Tracking Diffusion Module (PTDM), which are dedicated to high-quality eye texture generation and synchronized tracking of dynamic poses and textures, respectively. Compared to traditional GAN and VAE methods, ATGT3D significantly enhances texture consistency and generation quality in animated scenes using the EDM, which produces high-quality full-body textures with detailed eye information using the HUMBI dataset. Additionally, the Pose Tracking and Diffusion Module (PTDM) monitors human motion parameters utilizing the BEAT2 and AMASS mesh-level animatable human model datasets. The EDM, in conjunction with a basic texture seed featuring eyes and the diffusion model, restores high-quality textures, whereas the PTDM, by integrating MoSh++ and SMPL-X body parameters, models hand and body movements from 2D human images, thus providing superior 3D motion capture datasets. This module maintains the synchronization of textures and movements over time to ensure precise animation texture tracking. During training, the ATGT3D model uses the diffusion model as the generative backbone to produce new samples. The EDM improves the texture generation process by enhancing the precision of eye details in texture images. The PTDM involves joint training for pose generation and animation tracking reconstruction. Textures and body movements are generated individually using encoded prompts derived from masked gestures. Furthermore, ATGT3D adaptively integrates texture and animation features using the diffusion model to enhance both fidelity and diversity. Experimental results show that ATGT3D achieves optimal texture generation performance and can flexibly integrate predefined spatiotemporal animation inputs to create comprehensive human animation models. Our experiments yielded unexpectedly positive outcomes. Full article

Despite considerable advancements in integrating the Internet of Things (IoT) and artificial intelligence (AI) within the industrial maintenance framework, the increasing reliance on these innovative technologies introduces significant vulnerabilities due to cybersecurity risks, potentially compromising the integrity of decision-making processes. Accordingly, this study aims to offer comprehensive insights into the cybersecurity challenges associated with predictive maintenance, proposing a novel methodology that leverages generative AI for data augmentation, enhancing threat detection capabilities. Experimental evaluations conducted using the NASA Commercial Modular Aero-Propulsion System Simulation (N-CMAPSS) dataset affirm the viability of this approach leveraging the state-of-the-art TimeGAN model for temporal-aware data generation and building a recurrent classifier for attack discrimination in a balanced dataset. The classifier’s results demonstrate the satisfactory and robust performance achieved in terms of accuracy (between 80% and 90%) and how the strategic generation of data can effectively bolster the resilience of intelligent maintenance systems against cyber threats. Full article

To implement fine-grained progression signal control on arterial, it is essential to have access to the time-varying distribution of the origin–destination (OD) flow of the arterial. However, due to the sparsity of automatic vehicle identification (AVI) devices and the low penetration of connected vehicles (CVs), it is difficult to directly obtain the distribution pattern of arterial OD flow (i.e., path flow). To solve this problem, this paper develops a semi-supervised arterial path flow estimation method considering the consistency of path flow distribution by combining the sparse AVI data and the low permeability CV data. Firstly, this paper proposes a semi-supervised arterial path flow estimation model based on multi-knowledge graphs. It utilizes graph neural networks to combine some arterial AVI OD flow observation information with CV trajectory information to infer the path flow of AVI unobserved OD pairs. Further, to ensure that the estimation results of the multi-knowledge graph path flow estimation model are consistent with the distribution of path flow in real situations, we introduce a generative adversarial network (GAN) architecture to correct the estimation results. The proposed model is extensively tested based on a real signalized arterial. The results show that the proposed model is still able to achieve reliable estimation results under low connected vehicle penetration and with less observed label data. Full article

Assessing leakage risks in water distribution networks (WDNs) and implementing preventive monitoring for high-risk pipelines has become a widely accepted approach for leakage control. However, existing methods face significant data barriers between Geographic Information System (GIS) and leakage prediction systems. These barriers hinder traditional pipeline risk assessment methods, particularly when addressing challenges such as data imbalance, poor model interpretability, and lack of intuitive prediction results. To overcome these limitations, this study proposes a leakage assessment framework for water distribution networks based on multiple attention mechanisms and a generative model-based data balancing method. Extensive comparative experiments were conducted using water distribution network data from B2 and B3 District Metered Areas in Zhengzhou. The results show that the proposed model, optimized with a balanced data method, achieved a 40.76% improvement in the recall rate for leakage segment assessments, outperforming the second-best model using the same strategy by 1.7%. Furthermore, the strategy effectively enhanced the performance of all models, further proving that incorporating more valid data contributes to improved assessment results. This study comprehensively demonstrates the application of data-driven models in the field of “smart water management”, providing practical guidance and reference cases for advancing the development of intelligent water infrastructure. Full article

Atmospheric visibility is a crucial meteorological element impacting urban air pollution monitoring, public transportation, and military security. Traditional visibility detection methods, primarily manual and instrumental, have been costly and imprecise. With advancements in data science and computing, deep learning-based visibility detection technologies have rapidly emerged as a research hotspot in atmospheric science. This paper systematically reviews the applications of various deep learning models—Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Generative Adversarial Networks (GANs), and Transformer networks—in visibility estimation, prediction, and enhancement. Each model’s characteristics and application methods are discussed, highlighting the efficiency of CNNs in spatial feature extraction, RNNs in temporal tracking, GANs in image restoration, and Transformers in capturing long-range dependencies. Furthermore, the paper addresses critical challenges in the field, including dataset quality, algorithm optimization, and practical application barriers, proposing future research directions, such as the development of large-scale, accurately labeled datasets, innovative learning strategies, and enhanced model interpretability. These findings highlight the potential of deep learning in enhancing atmospheric visibility detection techniques, providing valuable insights into the literature and contributing to advances in the field of meteorological observation and public safety. Full article

This paper proposes DigitalUpSkilling, a novel IoT- and AI-based framework for improving and personalising the training of workers who are involved in physical-labour-intensive jobs. DigitalUpSkilling uses wearable IoT sensors to observe how individuals perform work activities. Such sensor observations are continuously processed to synthesise an avatar-like kinematic model for each worker who is being trained, referred to as the worker’s digital twins. The framework incorporates novel work activity recognition using generative adversarial network (GAN) and machine learning (ML) models for recognising the types and sequences of work activities by analysing an individual’s kinematic model. Finally, the development of skill proficiency ML is proposed to evaluate each trainee’s proficiency in work activities and the overall task. To illustrate DigitalUpSkilling from wearable IoT-sensor-driven kinematic models to GAN-ML models for work activity recognition and skill proficiency assessment, the paper presents a comprehensive study on how specific meat processing activities in a real-world work environment can be recognised and assessed. In the study, DigitalUpSkilling achieved 99% accuracy in recognising specific work activities performed by meat workers. The study also presents an evaluation of the proficiency of workers by comparing kinematic data from trainees performing work activities. The proposed DigitalUpSkilling framework lays the foundation for next-generation digital personalised training. Full article

The Sichuan Basin has abundant deep and ultra-deep natural gas resources, making it a primary target for exploration and the development of China’s oil and gas industry. However, during the drilling of ultra-deep wells in the Sichuan Basin, complex geological conditions frequently lead to gas kicks, posing significant challenges to well control and safety. Compared to traditional kick detection methods, artificial intelligence technology can improve the accuracy and timeliness of kick detection. However, there are limited real-world kick data available from drilling operations, and the datasets are extremely imbalanced, making it difficult to train intelligent models with sufficient accuracy and generalization capabilities. To address this issue, this paper proposes a kick data augmentation method based on a time-series generative adversarial network (TimeGAN). This method generates synthetic kick samples from real datasets and then employs a long short-term memory (LSTM) neural network to extract multivariate time-series features of surface drilling parameters. A multilayer perceptron (MLP) network is used for data classification tasks, constructing an intelligent kick detection model. Using real drilling data from ultra-deep wells in the SY block of the Sichuan Basin, the effects of k-fold cross-validation, data dimensionality, various imbalanced data handling techniques, and the sample imbalance ratio on the model’s kick detection performance are analyzed. Ablation experiments are also conducted to assess the contribution of each module in identifying kick. The results show that TimeGAN outperforms other imbalanced data handling techniques. The accuracy, recall, precision, and F1-score of the kick identification model are highest when the sample imbalance ratio is at 1 but decrease as the imbalance ratio increases. This indicates that maintaining a balance between positive and negative samples is essential for training a reliable intelligent kick detection model. The trained model is applied during the drilling of seven ultra-deep wells in Sichuan, demonstrating its effectiveness and accuracy in real-world kick detection. Full article

Due to its distinctive structure and unique physicochemical properties, gallium nitride (GaN) has been considered a prospective candidate for lithium storage materials. However, its inferior conductivity and unsatisfactory cycle performance hinder the further application of GaN as a next-generation anode material for lithium-ion batteries (LIBs). To address this, cobalt (Co)-doped GaN (Co-GaN) nanowires have been designed and synthesized by utilizing the chemical vapor deposition (CVD) strategy. The structural characterizations indicate that the doped Co elements in the GaN nanowires exist as Co²⁺ rather than metallic Co. The Co²⁺ prominently promotes electrical conductivity and ion transfer efficiency in GaN. The cycling capacity of Co-GaN reached up to 495.1 mA h g⁻¹ after 100 cycles. After 500 cycles at 10 A g⁻¹, excellent cycling capacity remained at 276.6 mA h g⁻¹. The intimate contact between Co-GaN nanowires and carbon paper enhances the conductivity of the composite. Density functional theory (DFT) calculations further illustrated that Co substitution changed the electron configuration in the GaN, which led to enhancement of the electron transfer efficiency and a reduction in the ion diffusion barrier on the Co-GaN electrode. This doping design boosts the lithium-ion storage performance of GaN as an advanced material in lithium-ion battery anodes and in other electrochemical applications. Full article

In actual production processes, analysis and prediction tasks commonly rely on large amounts of time-series data. However, real-world scenarios often face issues such as insufficient or imbalanced data, severely impacting the accuracy of analysis and predictions. To address this challenge, this paper proposes a dual-layer transfer model based on Generative Adversarial Networks (GANs) aiming to enhance the training speed and generation quality of time-series data augmentation under small-sample conditions while reducing the reliance on large training datasets. This method introduces a module transfer strategy based on the traditional GAN framework which balances the training between the discriminator and the generator, thereby improving the model’s performance and convergence speed. By employing a dual-layer network structure to transfer the features of time-series signals, the model effectively reduces the generation of noise and other irrelevant features, improving the similarity of the generated signals’ characteristics. This paper uses speech signals as a case study, addressing scenarios where speech data are difficult to collect and the limited number of speech samples available for effective feature extraction and analysis. Simulated speech timbre generation is conducted, and the experimental results on the CMU-ARCTIC database show that, compared to traditional methods, this approach achieves significant improvements in enhancing the consistency of generated signal features and the model’s convergence speed. Full article

The influence of environmental noise is generally excluded during research on machine fault diagnosis using acoustic signals. This study proposes a fault diagnosis method using a variational autoencoder (VAE) and domain adaptation neural network (DANN), both of which are based on unsupervised learning, to address this problem. The proposed method minimizes the impact of environmental noise and maintains the fault diagnosis performance in altered environments. The fault diagnosis algorithm was implemented using acoustic signals containing noise, present in the malfunctioning industrial machine investigation and inspection open dataset, and the fault prediction performance in noisy environments was examined based on forklift acoustic data using the VAE and DANN. The VAE primarily learns from normal state acoustic data and determines the occurrence of faults based on reconstruction error. To achieve this, statistical features of Mel frequency cepstral coefficients were extracted, generating features applicable regardless of signal length. Additionally, features were enhanced by applying noise reduction techniques via magnitude spectral subtraction and feature optimization, reflecting the characteristics of rotating equipment. Furthermore, data were augmented using generative adversarial networks to prevent overfitting. Given that the forklift acoustic data possess time-series characteristics, the exponentially weighted moving average was determined to quantitatively track time-series changes and identify early signs of faults. The VAE defined the reconstruction error as the fault index, diagnosing the fault states and demonstrating excellent performance using time-series data. However, the fault diagnosis performance of the VAE tended to decrease in noisy environments. Moreover, applying DANN for fault diagnosis significantly improved diagnostic performance in noisy environments by overcoming environmental differences between the source and target domains. In particular, by adapting the model learned in the source domain to the target domain and considering the domain differences based on signal-to-noise ratio, high diagnostic accuracy was maintained regardless of the noise levels. The DANN evaluated interdomain similarity using cosine similarity, enabling the accurate classification of fault states in the target domain. Ultimately, the combination of the VAE and DANN techniques enabled effective fault diagnosis even in noisy environments. Full article

With the rapid development of stereoscopic display technology, how to generate high-quality virtual view images has become the key in the applications of 3D video, 3D TV and virtual reality. The traditional virtual view rendering technology maps the reference view into the virtual view by means of 3D transformation, but when the background area is occluded by the foreground object, the content of the occluded area cannot be inferred. To solve this problem, we propose a virtual view acquisition technique for complex scenes of monocular images based on a layered depth image (LDI). Firstly, the depth discontinuities of the edge of the occluded area are reasonably grouped by using the multilayer representation of the LDI, and the depth edge of the occluded area is inpainted by the edge inpainting network. Then, the generative adversarial network (GAN) is used to fill the information of color and depth in the occluded area, and the inpainting virtual view is generated. Finally, GAN is used to optimize the color and depth of the virtual view, and the high-quality virtual view is generated. The effectiveness of the proposed method is proved by experiments, and it is also applicable to complex scenes. Full article