Search Results (27)

Fingerprint recognition systems have relied on fragile workflows based on minutiae extraction, which suffer from significant performance losses under real-world conditions such as sensor diversity and low image quality. This study introduces a fully minutiae-free fingerprint recognition framework based on self-supervised Vision Transformers. A systematic evaluation of multiple DINOv2 model variants is conducted, and the proposed system ultimately adopts the DINOv2-Base Vision Transformer as the primary configuration, as it offers the best generalization performance trade-off under conditions of limited fingerprint data. Larger variants are additionally analyzed to assess scalability and capacity limits. The DINOv2 pretrained network is fine-tuned using self-supervised domain adaptation on 64,801 fingerprint images, eliminating all classical enhancement, binarization, and minutiae extraction steps. Unlike the single-sensor protocols commonly used in the literature, the proposed approach is extensively evaluated in a heterogeneous testbed with a wide range of sensors, qualities, and acquisition methods, including 1631 unique fingers from 12 datasets. The achieved EER of 5.56% under these challenging conditions demonstrates clear cross-sensor superiority over traditional systems such as VeriFinger (26.90%) and SourceAFIS (41.95%) on the same testbed. A systematic comparison of different model capacities shows that moderate-scale ViT models provide optimal generalization under limited-data conditions. Explainability analyses indicate that the attention maps of the model trained without any minutiae information exhibit meaningful overlap with classical structural regions (IoU = 0.41 ± 0.07). Openly sharing the full implementation and evaluation infrastructure makes the study reproducible and provides a standardized benchmark for future research. Full article

Current object detection methods deployed in closed-circuit television (CCTV) systems experience substantial performance degradation due to domain gaps between training datasets and real-world environments. At the same time, increasing privacy concerns and stricter personal data regulations limit the reuse or distribution of source-domain data, highlighting the need for source-free learning. To address these challenges, we propose a stable and effective source-free semi-supervised domain adaptation framework based on the Mean Teacher paradigm. The method integrates three key components: (1) pseudo-label fusion, which combines predictions from weakly and strongly augmented views to generate more reliable pseudo-labels; (2) static adversarial regularization (SAR), which replaces dynamic discriminator optimization with a frozen adversarial head to provide a stable domain-invariance constraint; and (3) a time-varying exponential weighting strategy that balances the contributions of labeled and unlabeled target data throughout training. We evaluate the method on four benchmark scenarios: Cityscapes, Foggy Cityscapes, Sim10k, and a real-world CCTV dataset. The experimental results demonstrate that the proposed method improves mAP@0.5 by an average of 7.2% over existing methods and achieves a 6.8% gain in a low-label setting with only 2% labeled target data. Under challenging domain shifts such as clear-to-foggy adaptation and synthetic-to-real transfer, our method yields an average improvement of 5.4%, confirming its effectiveness and practical relevance for real-world CCTV object detection under domain shift and privacy constraints. Full article

This study adapts Fourier ptychography (FP) for high-resolution imaging in machine vision settings. We replace multi-angle illumination hardware with a single fixed light source and controlled object translation to enable a sequence of slightly shifted low-resolution frames to produce the requisite frequency-domain diversity for FP. The concept is validated in simulation using an embedded pupil function recovery algorithm to reconstruct a high-resolution complex field, recovering both amplitude and phase. For conveyor-belt transport, we introduce a lightweight preprocessing pipeline—background estimation, difference-based foreground detection, and morphological refinement—that yields robust masks and cropped inputs suitable for FP updates. The reconstructed images exhibit sharper fine structures and enhanced contrast relative to native lens imagery, indicating effective pupil synthesis without multi-LED arrays. The approach preserves compatibility with standard industrial optics and conveyor-style acquisition while reducing hardware complexity. We also discuss practical operating considerations, including blur-free capture and synchronization strategies. Full article

Batch normalization (BN) has become a foundational component in modern deep neural networks. However, one of its disadvantages is its reliance on batch statistics that may be unreliable or unavailable during inference, particularly under test-time domain shifts. While batch-statistics-free affine transformation methods alleviate this by learning per-sample scale and shift parameters, most treat samples independently, overlooking temporal or sequential correlations in streaming or episodic test-time settings. We propose LSTM-Affine, a memory-based normalization module that replaces BN with a recurrent parameter generator. By leveraging an LSTM, the module produces channel-wise affine parameters conditioned on both the current input and its historical context, enabling gradual adaptation to evolving feature distributions. Unlike conventional batch-statistics-free designs, LSTM-Affine captures dependencies across consecutive samples, improving stability and convergence in scenarios with gradual distribution shifts. Extensive experiments on few-shot learning and source-free domain adaptation benchmarks demonstrate that LSTM-Affine consistently outperforms BN and prior batch-statistics-free baselines, particularly when adaptation data are scarce or non-stationary. Full article

Aiming at the problems of high complexity in source domain data, inaccessibility of target domain data, and unknown fault patterns in real-world industrial scenarios for mechanical fault diagnosis, this paper proposes a Feature Decomposition-based Source-Free Universal Domain Adaptation (FD-SFUniDA) framework for mechanical equipment fault diagnosis. First, the CBAM attention module is incorporated to enhance the ResNet-50 convolutional network for extracting feature information from source domain data. During the target domain adaptation phase, singular value decomposition is applied to the weights of the pre-trained model’s classification layer, orthogonally decoupling the feature space into a source-known subspace and a target-private subspace. Then, based on the magnitude of feature projections, a dynamic decision boundary is constructed and combined with an entropy threshold mechanism to accurately distinguish between known and unknown class samples. Furthermore, intra-class feature consistency is strengthened through neighborhood-expanded contrastive learning, and semantic weight calibration is employed to reconstruct the feature space, thereby suppressing the negative transfer effect. Finally, extensive experiments under multiple operating conditions on rolling bearing and reciprocating mechanism datasets demonstrate that the proposed method excels in addressing source-free fault diagnosis problems for mechanical equipment and shows promising potential for practical engineering applications in fault classification tasks. Full article

Medical image segmentation faces significant challenges due to domain shift between different clinical centers and data privacy restrictions. Current source-free domain adaptation methods for medical images suffer from three critical limitations, including unstable training caused by noisy pseudo-labels and poor handling of foreground-background imbalance where critical structures like optic cup occupy extremely small regions. Additionally, strict privacy regulations often prevent access to source domain data during adaptation. To address these limitations, this paper proposes a source-free domain adaptation approach based on mutual information optimization for fundus image segmentation. The method incorporates a teacher–student network to ensure training stability and a mutual information maximization algorithm to reduce pseudo-label noise naturally. Furthermore, a prediction bank is constructed to handle class imbalance by leveraging complete statistics. Experimental results on fundus segmentation datasets demonstrate better performance, achieving 91.74% average Dice coefficient on Drishti-GS and 87.80% on RIM-ONE-r datasets, outperforming current methods. This work provides a practical solution for cross-institutional medical image analysis while preserving data privacy, with significant potential for eye disease diagnosis and other medical applications requiring robust domain adaptation. Full article

In tool wear monitoring, the efficient fusion of multi-source sensor signals poses significant challenges due to their inherent heterogeneous characteristics. In this paper, we propose a Multi-Sensor Multi-Domain feature fusion Transformer (MSMDT) model that achieves precise tool wear prediction through innovative feature engineering and cross-modal self-attention mechanisms. Specifically, we first develop a physics-aware feature extraction framework, where time-domain statistical features, frequency-domain energy features, and wavelet packet time–frequency features are systematically extracted for each sensor type. This approach constructs a unified feature matrix that effectively integrates the complementary characteristics of heterogeneous signals while preserving discriminative tool wear signatures. Then, a position-embedding-free Transformer architecture is constructed, which enables adaptive cross-domain feature fusion through joint global context modeling and local feature interaction analysis to predict tool wear values. Experimental results on the PHM2010 demonstrate the superior performance of MSMDT, outperforming state-of-the-art methods in prediction accuracy. Full article

Modern vehicles are rapidly transforming into interconnected cyber–physical systems that rely on advanced sensor technologies and pervasive connectivity to support autonomous functionality. Yet, despite this evolution, standardized methods for quantifying cybersecurity vulnerabilities across critical automotive components remain scarce. This paper introduces a novel hybrid model that integrates expert-driven insights with generative AI tools to adapt and extend the Common Vulnerability Scoring System (CVSS) specifically for autonomous vehicle sensor systems. Following a three-phase methodology, the study conducted a systematic review of 16 peer-reviewed sources (2018–2024), applied CVSS version 4.0 scoring to 15 representative attack types, and evaluated four free source generative AI models—ChatGPT, DeepSeek, Gemini, and Copilot—on a dataset of 117 annotated automotive-related vulnerabilities. Expert validation from 10 domain professionals reveals that Light Detection and Ranging (LiDAR) sensors are the most vulnerable (9 distinct attack types), followed by Radio Detection And Ranging (radar) (8) and ultrasonic (6). Network-based attacks dominate (104 of 117 cases), with 92.3% of the dataset exhibiting low attack complexity and 82.9% requiring no user interaction. The most severe attack vectors, as scored by experts using CVSS, include eavesdropping (7.19), Sybil attacks (6.76), and replay attacks (6.35). Evaluation of large language models (LLMs) showed that DeepSeek achieved an F1 score of 99.07% on network-based attacks, while all models struggled with minority classes such as high complexity (e.g., ChatGPT F1 = 0%, Gemini F1 = 15.38%). The findings highlight the potential of integrating expert insight with AI efficiency to deliver more scalable and accurate vulnerability assessments for modern vehicular systems.This study offers actionable insights for vehicle manufacturers and cybersecurity practitioners, aiming to inform strategic efforts to fortify sensor integrity, optimize network resilience, and ultimately enhance the cybersecurity posture of next-generation autonomous vehicles. Full article

Radio frequency fingerprint identification (RFFI) leverages the unique characteristics of radio signals resulting from inherent hardware imperfections for identification, making it essential for applications in telecommunications, cybersecurity, and surveillance. Despite the advancements brought by deep learning in enhancing RFFI accuracy, challenges persist in model deployment, particularly when transferring RFFI models across different receivers. Variations in receiver hardware can lead to significant performance declines due to shifts in data distribution. This paper introduces the source-free cross-receiver RFFI (SCRFFI) problem, which centers on adapting pre-trained RF fingerprinting models to new receivers without needing access to original training data from other devices, addressing concerns of data privacy and transmission limitations. We propose a novel approach called contrastive source-free cross-receiver network (CSCNet), which employs contrastive learning to facilitate model adaptation using only unlabeled data from the deployed receiver. By incorporating a three-pronged loss function strategy—minimizing information entropy loss, implementing pseudo-label self-supervised loss, and leveraging contrastive learning loss—CSCNet effectively captures the relationships between signal samples, enhancing recognition accuracy and robustness, thereby directly mitigating the impact of receiver variations and the absence of source data. Our theoretical analysis provides a solid foundation for the generalization performance of SCRFFI, which is corroborated by extensive experiments on real-world datasets, where under realistic noise and channel conditions, that CSCNet significantly improves recognition accuracy and robustness, achieving an average improvement of at least 13% over existing methods and, notably, a 47% increase in specific challenging cross-receiver adaptation tasks. Full article

Intelligent fault diagnosis for rotary machinery often suffers performance degradation under domain shifts between training and deployment environments. To address this, we propose a robust fault diagnosis framework incorporating three key components: (1) an order-frequency-based preprocessing method to normalize rotational variations, (2) a U-Net variational autoencoder (U-NetVAE) to enhance adaptation through reconstruction learning, and (3) a test-time training (TTT) strategy enabling unsupervised target domain adaptation without access to source data. Since existing works rarely evaluate under true domain shift conditions, we first construct a unified cross-domain benchmark by integrating four public datasets with consistent class and sensor settings. The experimental results show that our method outperforms conventional machine learning and deep learning models in both F1-score and recall across domains. Notably, our approach maintains an F1-score of 0.47 and recall of 0.51 in the target domain, outperforming others under identical conditions. Ablation studies further confirm the contribution of each component to adaptation performance. This study highlights the effectiveness of combining mechanical priors, self-supervised learning, and lightweight adaptation strategies for robust fault diagnosis in the practical domain. Full article

Abdominal organ segmentation in CT images is crucial for accurate diagnosis, treatment planning, and condition monitoring. However, the annotation process is often hindered by challenges such as low contrast, artifacts, and complex organ structures. While unsupervised domain adaptation (UDA) has shown promise in addressing these issues by transferring knowledge from a different modality (source domain), its reliance on both source and target data during training presents a practical challenge in many clinical settings due to data privacy concerns. This study aims to develop a cross-modality abdominal multi-organ segmentation model for label-free CT (target domain) data, leveraging knowledge solely from a pre-trained source domain (MRI) model without accessing the source data. To achieve this, we generate source-like images from target-domain images using a one-way image translation approach with the pre-trained model. These synthesized images preserve the anatomical structure of the target, enabling segmentation predictions from the pre-trained model. To further enhance segmentation accuracy, particularly for organ boundaries and small contours, we introduce an auxiliary translation module with an image decoder and multi-level discriminator. The results demonstrate significant improvements across several performance metrics, including the Dice similarity coefficient (DSC) and average symmetric surface distance (ASSD), highlighting the effectiveness of the proposed method. Full article

Due to access to the source data during the transfer phase, conventional domain adaptation works have recently raised safety and privacy concerns. More research attention thus shifts to a more practical setting known as source-data-free domain adaptation (SFDA). The new challenge is how to obtain reliable semantic supervision in the absence of source domain training data and the labels on the target domain. To that end, in this work, we introduce a novel Gradual Geometry-Guided Knowledge Distillation (G2KD) approach for SFDA. Specifically, to address the lack of supervision, we used local geometry of data to construct a more credible probability distribution over the potential categories, termed geometry-guided knowledge. Then, knowledge distillation was adopted to integrate this extra information for boosting the adaptation. More specifically, first, we constructed a neighborhood geometry for any target data using a similarity comparison on the whole target dataset. Second, based on pre-obtained semantic estimation by clustering, we mined soft semantic representations expressing the geometry-guided knowledge by semantic fusion. Third, using the soften labels, we performed knowledge distillation regulated by the new objective. Considering the unsupervised setting of SFDA, in addition to the distillation loss and student loss, we introduced a mixed entropy regulator that minimized the entropy of individual data as well as maximized the mutual entropy with augmentation data to utilize neighbor relation. Our contribution is that, through local geometry discovery with semantic representation and self-knowledge distillation, the semantic information hidden in the local structures is transformed to effective semantic self-supervision. Also, our knowledge distillation works in a gradual way that is helpful to capture the dynamic variations in the local geometry, mitigating the previous guidance degradation and deviation at the same time. Extensive experiments on five challenging benchmarks confirmed the state-of-the-art performance of our method. Full article

Human–computer interaction (HCI) drives innovation by bridging humans and technology, with human activity recognition (HAR) playing a key role. Traditional HAR systems require user cooperation and infrastructure, raising privacy concerns. In recent years, Wi-Fi devices have leveraged channel state information (CSI) to decode human movements without additional infrastructure, preserving privacy. However, these systems struggle with unseen users, new environments, and scalability, thereby limiting real-world applications. Recent research has also demonstrated that the impact of surroundings causes dissimilar variations in the channel state information at different times of the day. In this paper, we propose an unsupervised multi-source domain adaptation technique that addresses these challenges. By aligning diverse data distributions with target domain variations (e.g., new users, environments, or atmospheric conditions), the method enhances system adaptability by leveraging public datasets with varying domain samples. Experiments on three public CSI datasets using a preprocessing module to convert CSI into image-like formats demonstrate significant improvements to baseline methods with an average micro-F1 score of 81% for cross-user, 76% for cross-user and cross-environment, and 73% for cross-atmospheric tasks. The approach proves effective for scalable, device-free sensing in realistic cross-domain HAR scenarios. Full article

The solution of the Helmholtz equation describing the propagation of light in free space from a plane to another can be described by the angular spectrum operator, which acts in the frequency domain. Many applications require this operator to be generalized to handle tilted source and target planes, which has led to research investigating the implications of these adaptations. However, the frequency domain representation intrinsically limits the understanding the way the signal is transformed through propagation. Instead, studying how the operator maps the space–frequency components of the wavefield provides essential information that is not available in the frequency domain. In this work, we highlight and exploit the deep relation between wave optics and quantum mechanics to explicitly describe the symplectic action of the tilted angular spectrum in phase space, using mathematical tools that have proven their efficiency for quantum particle physics. These derivations lead to new algorithms that open unprecedented perspectives in various domains involving the propagation of coherent light. Full article

Deep learning techniques for medical image analysis often encounter domain shifts between source and target data. Most existing approaches focus on unsupervised domain adaptation (UDA). However, in practical applications, many source domain data are often inaccessible due to issues such as privacy concerns. For instance, data from different hospitals exhibit domain shifts due to equipment discrepancies, and data from both domains cannot be accessed simultaneously because of privacy issues. This challenge, known as source-free UDA, limits the effectiveness of previous UDA medical methods. Despite the introduction of various medical source-free unsupervised domain adaptation (MSFUDA) methods, they tend to suffer from an over-fitting problem described as “longer training, worse performance”. To address this issue, we proposed the Stable Learning (SL) strategy. SL is a method that can be integrated with other approaches and consists of weight consolidation and entropy increase. Weight consolidation helps retain domain-invariant knowledge, while entropy increase prevents over-learning. We validated our strategy through experiments on three MSFUDA methods and two public datasets. For the abdominal dataset, the application of the SL strategy enables the MSFUDA method to effectively address the domain shift issue. This results in an improvement in the Dice coefficient from 0.5167 to 0.7006 for the adaptation from CT to MRI, and from 0.6474 to 0.7188 for the adaptation from MRI to CT. The same improvement is observed with the cardiac dataset. Additionally, we conducted ablation studies on the two involved modules, and the results demonstrated the effectiveness of the SL strategy. Full article