Sensors最新文献

{"title":"A Multiscale Electromagnetic Method for Damage Detection and Quantitative Characterization of Radar-Absorbing Coatings on Aircraft Skins.","authors":"Yang Fang, Wenbin Ren, Yijun Liu, Wei Cui, Yongli Ma, Bin Liu, Yong Li, Zhenmao Chen","doi":"10.3390/s26196357","DOIUrl":"10.3390/s26196357","url":null,"abstract":"<p><p>This study proposes a multiscale electromagnetic method that combines far-field angular-domain analysis with local near-field imaging to detect damage in radar-absorbing coatings on aircraft surfaces. Using an intact state as a reference, continuous anomalous angular intervals and their principal anomalous azimuths are identified via robust statistical criteria, employing the electromagnetic scattering energy ratio (ESER), defined here as a normalized squared-field response ratio, and the azimuth difference factor. These angular characteristics, combined with target geometry and a predefined detection area, are used to determine the direction and azimuthal range of candidate near-field scanning windows. Subsequently, coherent back-projection is applied to the local near-field measurements. The centroid and equivalent electromagnetic dimensions of the differential response are estimated using complex differential imaging, multithreshold analysis, and squared-magnitude-weighted central second moments. Numerical analysis of a simplified aircraft-like model demonstrates angular-domain screening and local characterization of impact, coating loss and bulging. For the rectangular coating-loss simulation, the equivalent electromagnetic length and width deviate from the nominal side lengths by 8.6% and 2.3%, respectively. The corresponding experimental deviations for localized 1 cm × 1 cm coating loss on a coated conical specimen are 9.2% and 6.3%. These results support two-stage screening and local equivalent-scale extraction under the investigated geometry, material and reference-matching conditions.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667452/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196367","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Newly Developed Tolles-Lawson Compensation Model Accounting for Eddy-Current Dynamic Response.","authors":"Peng Gu, Dong Gao, Chunjiang Bian, Minghui Zhu","doi":"10.3390/s26196338","DOIUrl":"10.3390/s26196338","url":null,"abstract":"<p><p>Aeromagnetic compensation mitigates the aircraft-induced magnetic interference in airborne magnetometer measurements, thereby providing the measurement accuracy required by geomagnetic navigation and aeromagnetic surveying. The most widely used compensation framework, the conventional Tolles-Lawson (T-L) model, approximates the eddy-current magnetic field as an instantaneous linear response to the rate of change of the external magnetic field. Our analysis demonstrates that this assumption becomes invalid when the structural characteristic time scale is comparable to the aircraft maneuver time scale. We therefore generalize the eddy-current term into a first-order dynamic element, resulting in a dynamic eddy-current T-L model in which the eddy-current response is no longer instantaneous, while the 18-term design-matrix structure is preserved and no additional parameters are introduced. Utilizing a public flight dataset, we compare the two models on independent validation flight lines with a single calibration segment. Ridge regression on the Mag 4 channel yields a consistent and statistically significant improvement, where the median enhancement in the high-dynamic flight lines reaches +1.177 nT, approximately 70 times greater than that observed in regular survey lines. This benefit arises precisely under the high-dynamic flight conditions that are most critical for geomagnetic navigation, and its robustness is confirmed by dual-reference, vector-input, and regularization-parameter analyses.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667723/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196379","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Fine-Tuning-Free Reference-Guided Diffusion for Structure-Preserving Elongated Fabric Defect Generation.","authors":"Qiang Wu, Hedong Yang, Wensheng Ding, Zhong Xiang, Jinling Yao","doi":"10.3390/s26196351","DOIUrl":"10.3390/s26196351","url":null,"abstract":"<p><p>In industrial anomaly detection, anomaly generation helps reduce reliance on scarce real defect samples. However, for elongated fabric defects where morphology is highly coupled with background texture, maintaining the realism of the defects remains a challenge. To address this issue, we propose the Shape-Aware Implicitly Anchored Industrial Defect Generation (SLAID) framework, a reference-image-guided synthesis framework for elongated fabric defects that is fine-tuning-free. Within this framework, we introduce Shape-Sensitive Sliding-Window Reference Modeling (SSRM) to preserve the geometric structure of long-span defects. Additionally, we incorporate Adaptive High-Frequency Guidance (AHFG) to enhance fine edges and textural features. Finally, we propose Target-Domain Latent Anchoring (TLA) to constrain target-background consistency in non-edited regions. SLAID generates structurally continuous defects while preserving the target fabric background, without additional fine-tuning. In a data augmentation experiment on a dataset comprising five categories of elongated textile defects, SLAID improved the mAP@0.5 of YOLOv8 from 0.616 to 0.817. Furthermore, SLAID achieved a mean MMD² of 0.0435 for defect-distribution consistency, a mean LPIPS of 0.0739 for defect-level perceptual similarity, and a full-image SSIM of 0.9754 for target-background preservation. Cross-dataset evaluation on the MHPSD steel-surface defect dataset provides preliminary evidence of the applicability of the proposed inference-stage generation strategy to the evaluated steel-surface defect categories.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667604/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196451","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Expert-Consistency-Guided Proposal Refinement for 3D Object Detection from Sparse Radar Point Clouds.","authors":"Zixuan Liu, Wei Xiong","doi":"10.3390/s26196339","DOIUrl":"10.3390/s26196339","url":null,"abstract":"<p><p>Sparse radar point clouds provide limited evidence for three-dimensional (3D) object detection, making weak proposals especially sensitive to ranking and suppression. Standard post-processing depends on the primary detector's scores and geometry and can therefore remove useful candidates. We introduce an expert-consistency-guided refinement framework that acts on detector outputs without changing the backbone. Candidate boxes are retained and reweighted according to agreement with an independently trained radar expert; supported boxes are then adjusted by local geometric voting. A separate high-performance route instead trains a localization-quality-aware confidence head with relaxed positive assignment and does not use the expert-consistency gate. The framework is evaluated on Astyx, MAN TruckScenes, V2X-Radar-V, and K-Radar with three training seeds. At a 3D intersection-over-union threshold of 0.50, the strict route improves average precision (AP) in all 12 matched dataset-seed comparisons, with a mean gain of 0.9726 AP. The high-performance route, whose gain comes from this quality-aware ranking branch rather than from expert-consistency gating, raises macro AP from 35.3552 to 39.1214, a gain of 3.7662 AP. These improvements are observations under the evaluated protocol: the strict voting parameters were selected using the same dataset-seed cells, and leave-one-dataset-out threshold selection yields 11/12 positive held-out differences. The strict route also shows metric-dependent confidence-calibration trade-offs and requires an additional expert detector and voting computation. The results therefore support protocol-specific gains, not a guarantee of improved reliability on unseen data.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667582/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149195413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A New Current Fingerprinting Sensing Method for Adaptive Protection of PV-Integrated Distribution Networks.","authors":"Ahmed A Salem, Emad Eldeen Omran, Azza A ElDesouky, Hany S E Mansour, Mohamed A A Mohamed, Eyad S Oda","doi":"10.3390/s26196323","DOIUrl":"10.3390/s26196323","url":null,"abstract":"<p><p>Grid-following photovoltaic (PV) inverters inject fault currents whose magnitude and phase are set by the control mode active during the disturbance, so directional overcurrent relay (DOCR) settings coordinated for one inverter response can miscoordinate after a mode change. This paper proposes a communication-free adaptive protection scheme built on local sensing. A voltage-adaptive current fingerprinting classifier uses local positive sequence measurements to detect current-limiting (CL), active current priority (ACP) and reactive current priority (RCP) operation, as well as momentary cessation (MC) and asymmetrical faults. The classifier activates the relay setting group prepared offline for the detected mode. With pickup currents fixed, the coordination sub-problem is shown to be a linear program (LP), so every setting group is computed at its exact global optimum. Where the grid code droop law makes ACP and RCP indistinguishable by angle, the classifier declares the ambiguity and activates a group that is feasible for both. Under a class-0.5 instrument transformer noise model, the group placed in service is coordination-valid in 99.8% of cases and the expected residual miscoordination is 3.0×10-5 s, 0.01% of the 0.30 s margin. On the CIGRE medium-voltage benchmark, the scheme satisfies the margin in all operating modes for three-phase, line-to-line (LL) and single-line-to-ground (SLG) faults. The same fault types are then verified at three fault locations on a practical Egyptian 11 kV feeder model, where the sensing layer is confirmed on an electromagnetic transient (EMT) model and the scheme removes the violations of single-setting coordination using an estimated 4 kB of relay memory.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667734/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196344","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Permutation-Invariant, Missing-View-Aware Multi-Perspective Recurrent Neural Network Cell for Time-Series Fusion.","authors":"Chao Han, Ruidong Wang, Lilong Zhang","doi":"10.3390/s26196330","DOIUrl":"10.3390/s26196330","url":null,"abstract":"<p><p>Recurrent models for multi-source data must combine synchronized streams while remaining usable when sources are reordered or unavailable. Multi-perspective long short-term memory (MP-LSTM) cells update a joint memory across perspectives, but a sequential update can make the result depend on source order. We implement a generalized chain comparator and replace its perspective-wise chain with shared projections and set attention over available views. All experiments use one fixed subject-independent split of the UCI Daily and Sports Activities (DSA) dataset, with subjects 7 and 8 as the two held-out test participants. On this split, the proposed cell uses 66.0% fewer parameters and has a 41.5% shorter recorded training time than the comparator. On a fixed 720-segment, seed-0 test subset, its inference-time accuracy remains 75.14% across eight random view permutations (standard deviation (SD) 0.00 percentage points), whereas the chain comparator falls from 76.39% to a mean of 24.67%. In the 10-class task, the mean full-view accuracy across three training seeds is 3.53 percentage points lower for the proposed cell. A cyclic missing-view schedule increases seed-0 accuracy with four of five views removed from 17.47% to 41.41% in the 19-class task, while three-seed mean full-view accuracy decreases by 3.40 percentage points. These results characterize order-invariant fusion and its robustness-accuracy trade-off in this controlled case study; they do not establish generalization beyond this dataset and held-out-subject split.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667684/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149195739","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"CyberRoBERTa: A Lightweight Transformer-Based Language Model for IoT Intrusion Detection.","authors":"Dhai Alhafi, Fatmah Alanazi","doi":"10.3390/s26196326","DOIUrl":"10.3390/s26196326","url":null,"abstract":"<p><p>With the accelerated proliferation of the Internet of Things (IoT) and its widespread use in sensitive areas such as healthcare, homes, and factories, security has become an increasingly critical concern. IoT remains one of the most challenging domains to secure, lacking the manageable scale, well-maintained software, and homogeneous environments of easily secured systems. Large language models (LLMs) have recently been applied to improve IoT security; however, most applications treat LLMs as \"off-the-shelf\" tools, creating a domain mismatch because model representations, tokenizers, and data collators are designed for natural language rather than network traffic. In this work, we present CyberRoBERTa, built specifically to test whether departing from off-the-shelf LLM components, through a domain-specific Unigram tokenizer and an IoT-specialized data collator, yields a genuine advantage for IoT intrusion detection. We instantiate this framework as a compact, 10.2-million-parameter model designed for fully self-contained, on-device operation. Using CICIoT2023, TON_IoT, and Edge-IIoTset, the model achieves a binary classification F1-score of 96% and a multiclass F1-score of 89.54%. We further validate these results in three ways: a comparison against a classical machine learning baseline shows comparable accuracy; isolating the tokenizer's contribution from pretraining scale shows its benefit holds under matched pretraining conditions; and evaluating class-imbalance interventions shows improved detection on the most data-scarce class.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667479/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196396","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Data-Minimizing Motion-Gated Artificial Intelligence Decision Support for Electric Micro-Mobility: An Architecture-Feasibility Study.","authors":"Syed Tahir Ali Shah, José Paulo Oliveira Santos, José Maria Fernandes, António B Pereira, Gabriel Constantinescu","doi":"10.3390/s26196334","DOIUrl":"10.3390/s26196334","url":null,"abstract":"<p><p>Physiological sensing can support electric micro-mobility safety research. Elevated heart rate during riding, however, may reflect physical exertion, psychological stress, motion artefacts, road context, or several of these factors. This paper presents RideSafe-AI, a non-clinical architecture-feasibility prototype. It separates data quality, exertion/load proxy estimation, public-dataset stress-label classification, and motion reliability before applying deterministic fusion and abstention. The Python reference checks input and output fields for raw physiological streams and direct identifiers, while the ESP32-S3 (Espressif Systems, Shanghai, China) checks incoming fields; Android constructs bounded inputs internally without an equivalent schema guard. PAMAP2 supports exertion/load proxy modelling, WESAD supports low/high laboratory condition-label classification, and official PhysioNet stress/exercise records support a subject-wise laboratory feature-set experiment. On the PhysioNet binary stress-versus-exercise task, full-feature XGBoost achieved 0.963 accuracy, 0.944 macro-F1, and a 0.015 exercise false-stress rate on 10,824 windows from 36 subjects. These results describe that public-dataset task; PhysioNet does not contain bike-frame motion, GPS terrain, or power telemetry and does not evaluate the complete RideSafe-AI pipeline. Deterministic scenarios verify selected decision rules, while two same-day technical rides by one participant demonstrate prototype integration and logging. These findings support architecture feasibility. Rider-stress accuracy, safety impact, and performance across riders and road conditions remain to be evaluated.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667601/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196423","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Lesion-Aware Multiple-Instance Learning for Grapevine Disease Recognition from Field RGB Images.","authors":"Övünç Öztürk, Bora Canbula, Tuğba Özacar, Vehbi Özacar","doi":"10.3390/s26196318","DOIUrl":"10.3390/s26196318","url":null,"abstract":"<p><p>This retrospective offline study examines how dense manual regional annotations can be organized for image-level grapevine disease recognition using pre-acquired HERMOS RGB images. After audit and filtering, 491 source images and 13,765 Pascal VOC regions were retained. A leakage-safe source-image split was used across full-image classification, regional crop classification, crop-to-image aggregation, annotation-guided multiple-instance learning (MIL), and YOLO11 localization. MIL bags in training, validation, and testing were constructed from crops derived from manual boxes; the results therefore describe annotation-assisted aggregation rather than raw-image deployment. The original ConvNeXt-Tiny MIL model achieved macro-F1 0.922 (95% source-image bootstrap CI 0.882-0.956), compared with 0.836 (0.754-0.895) for full-image ConvNeXt-Tiny and 0.869 (0.804-0.917) for the best fixed crop aggregation. Paired macro-F1 differences favored MIL over these baselines by 0.087 (95% CI 0.029-0.163; Holm-adjusted <i>p</i> = 0.013) and 0.053 (0.005-0.114; adjusted <i>p</i> = 0.036), respectively. Downy mildew precision and recall were 1.000 on nine positive test images, but the corresponding exact 95% binomial intervals were 0.664-1.000. Controlled ablations produced lower point estimates without learned attention or the auxiliary instance loss, whereas removing context expansion did not reduce performance; none of the three paired ablation tests was significant after Holm correction. Increasing detector capacity from YOLO11n to YOLO11s did not improve full-image mAP@0.50 (0.175 versus 0.170). These findings support annotation-guided local-evidence aggregation as an offline exploratory benchmark, while not establishing split-to-split stability, computational efficiency, or external-vineyard generalization.</p>","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667579/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196487","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Fine-Grained Image Dataset of Gold and Silver Artifacts from the Tomb of Prince Liangzhuang and Preliminary LoRA-Based Generation Experiments.","authors":"Yelei Chen, Yifan Xu, Lei Shu, Ru Han","doi":"10.3390/s26196322","DOIUrl":"10.3390/s26196322","url":null,"abstract":"&lt;p&gt;&lt;p&gt;The digital preservation of cultural heritage is evolving from static archiving towards intelligent creation. As representative examples of early Ming Dynasty gold and silver craftsmanship, the gold and silver artifacts from the Tomb of Prince Liangzhuang, with their complex craftsmanship and rich cultural significance embodied in their decorative motifs, pose substantial challenges for digital preservation and innovative design. Generative artificial intelligence currently faces difficulties in reproducing fine-grained forms and decorative features, while the lack of high-quality, multi-dimensionally annotated image data specifically designed for such artifacts further constrains their digital preservation and application. To address these issues, this study constructed a fine-grained, multi-dimensional annotated image dataset of gold and silver artifacts from the Tomb of Prince Liangzhuang. The dataset covers five major categories, namely headwear, headgear and sashes, accessories, daily necessities, and ritual objects, comprising 82 artifacts and 4449 images, including 4127 panoramic images and 322 detail images. All images were annotated according to four visual feature dimensions: form and structure, decorative patterns, material appearance, and color composition. Based on this dataset, LoRA fine-tuning was performed on three mainstream diffusion models, namely SD1.5, SDXL, and FLUX. In the experimental evaluation, a total of 1968 images were generated for the 82 artifacts using the three models and eight random seeds (82 artifacts × 3 models × 8 random seeds). The generation results were systematically evaluated using quantitative metrics, including Structural Similarity Index Measure (SSIM), Learned Perceptual Image Patch Similarity (LPIPS), CLIP Score, and Kernel Inception Distance (KID), together with five types of subjective evaluation questionnaires, yielding 101 valid questionnaire responses in total. The results showed that all three LoRA-fine-tuned models were capable of generating images exhibiting typical characteristics of the gold and silver artifacts from the Tomb of Prince Liangzhuang. SD1.5_LoRA achieved the highest SSIM score (0.627) and the lowest seed-to-seed standard deviation (0.050), with statistically significant differences observed in most pairwise comparisons across categories after Bonferroni correction. SDXL_LoRA achieved the highest composite score in the subjective evaluation. CLIP Score showed category-dependent differences across the five artifact categories, and no consistent overall ranking of the three models was observed, suggesting comparable semantic alignment performance across the models. The resolution comparison experiment showed that, for the headwear category using SDXL_LoRA, no significant differences were observed across the evaluated metrics between the 768 × 768 and 1024 × 1024 resolutions. The results demonstrate that the proposed dataset can effectively support LoRA fine-tuning for the ","PeriodicalId":21698,"journal":{"name":"Sensors","volume":"26 19","pages":""},"PeriodicalIF":4.0,"publicationDate":"2026-10-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13667451/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"149196254","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"综合性期刊","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}