Search Results (997)

Urban planners need scalable ways to monitor pedestrian conditions across heterogeneous cities, but conventional Level-of-Service (LOS) methods are often locally calibrated and difficult to transfer. This study proposes a city-adaptive framework for pedestrian LOS mapping using spatio-temporal graph models and few-shot transfer learning. Pedestrian count data from Melbourne, Dublin, and Zurich were converted into six ordinal LOS classes using city-specific percentile thresholds computed from the training data, yielding a relative congestion measure rather than an absolute cross-city standard. We developed a spatio-temporal graph transformer with an ordinal prediction head and evaluated it under in-domain, zero-shot, few-shot, and domain-adaptive settings. The results show strong in-domain performance in Melbourne (accuracy 79.7%; Acc ± 1 99.1%) and effective adaptation to the city-adaptive ordinal classification task. Few-shot fine-tuning with only 5% labeled target city data recovered 95–99% of in-domain performance, suggesting that small amounts of local supervision can substantially reduce calibration requirements in data-scarce environments. KernelSHAP analysis indicates that short-term temporal lag features dominate predictions across cities, whereas spatial and contextual features vary more strongly with local urban structure. The findings suggest that few-shot transfer learning can support pedestrian LOS estimation in cities with limited labeled data; however, the proposed LOS formulation should be interpreted as a city-specific relative indicator rather than an absolute measure of pedestrian comfort, crowding, or service quality. While the framework was evaluated across three cities, additional validation in diverse urban contexts and against perceptual measures of pedestrian experience remains necessary. Overall, the study contributes a city-adaptive framework for transferable relative LOS prediction rather than a universal cross-city LOS standard. Full article

Accurate power load forecasting is essential for efficient electricity grid management, yet capturing cross-variate dependencies in multivariate time series remains a persistent challenge. Recent channel-independent methods based on Transformer and recurrent architectures have achieved strong forecasting performance, but they discard potentially useful information from correlated variates such as weather conditions and neighboring consumption zones. In this paper, we propose RWKV-CVM, a lightweight extension of the RWKV-TS architecture that introduces a trainable Cross-Variate Mixing (CVM) module to selectively incorporate inter-variate information while preserving the linear time complexity of the backbone. The CVM module is a gated, row-stochastic mixing matrix—initialized from the training set absolute Pearson correlations and modulated by a single learned scalar gate that is applied to the normalized input series before patching, adding only 65 trainable parameters to the backbone. We evaluate the method under a single unified harness (three random seeds, consistent normalization, and re-executed DLinear, iTransformer and RWKV-TS baselines) on three settings: the Tetouan city power consumption dataset forecast jointly for all three zones at horizons up to 72 h (including the operationally relevant 24 h day-ahead and 48 h two-day-ahead horizons) and the ETTh1 and Weather benchmarks under a  10 %  few-shot protocol. Averaged over horizons, RWKV-CVM attains the lowest mean MSE on all three datasets (Tetouan all-zone  0 . 0427 , ETTh1  0 . 640 , Weather  0 . 250 ), narrowly ahead of the strongly-tuned baselines and its own RWKV-TS backbone. The advantage is modest, is concentrated at longer horizons, and is selective across target zones; on several individual horizons and in the full-data regime, a baseline is preferable, and we report these cases explicitly. These results indicate that a controlled, lightweight injection of cross-variate information can improve multivariate load forecasting on average without sacrificing computational efficiency. Full article

Enterprise financial risk prediction is a fundamental task in financial risk management, yet its performance is often hindered by severe class imbalance, cross-enterprise heterogeneity, and the limited availability of labeled risky samples. These challenges are particularly pronounced in few-shot settings, where conventional machine learning and deep classification models tend to suffer from unstable representation learning, feature collapse, and weak decision boundaries. To address this issue, this study proposes a hierarchical metric learning framework for few-shot enterprise financial risk prediction on imbalanced tabular data. The framework integrates a state-space feature embedding network, an Adaptive Spectral Decomposition and Multi-Scale State Embedding module, and a Hierarchical Metric Manifold Alignment mechanism to enhance risk-sensitive representation learning, preserve geometric consistency across embedding levels, and improve prototype-based discrimination in the metric space. Experiments are conducted on three public datasets, namely American Bankruptcy, Corporate Financial Risk Assessment, and Enterprise Financial Network, under a unified 2-way 20-shot setting. The proposed method consistently achieves the best overall performance across Precision, Recall, Accuracy, F1-score, and AUC, with AUC values of 0.9526, 0.9687, and 0.9716 on the three datasets, respectively. Ablation studies and visual analyses further show that the proposed framework improves intra-class compactness, inter-class separability, and classification robustness under highly imbalanced conditions. These findings indicate that the proposed method provides an effective and robust machine learning solution for enterprise financial risk prediction and early warning in data-scarce financial scenarios. Full article

Class-incremental learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. Recent advances in parameter-efficient fine-tuning (PEFT) based on pre-trained models (PTMs) have shown promise in this setting by integrating new tasks with minimal parameter overhead. However, these methods often suffer from knowledge degradationdue to: (1) cumulative interference caused by iterative updates, constrained gradient flows, or entangled module integration; and (2) suboptimal alignment between inference samples and specialized modules. To address these challenges, we propose Dynamic LoRA-Experts and Prototype-Ensemble Matching (DLEPEM), a novel two-stage, rehearsal-free framework. In the first stage, we allocate a task-specific LoRA-Expert for each incremental task, enabling isolated representation learning and reducing cross-task interference. In the second stage, we introduce a prototype-ensemble-matching mechanism that combines general prototypes derived from the frozen PTM with task-adaptive prototypes learned by the LoRA-Experts. This fusion facilitates both strong generalization and precise task-level discrimination. Extensive experiments on standard CIL and few-shot class-incremental learning (FSCIL) benchmarks demonstrate that DLEPEM achieves strong performance under the evaluated protocols. For instance, in CIL, it achieves 93.39% on CIFAR100 (+0.80% over EASE), 92.31% on CUB200 (+2.11% over EASE), and 91.84% on VTAB (+1.39% over EASE). In the more challenging FSCIL setting, it achieves 88.77% on CUB200, outperforming the strongest baseline by a clear margin of 5.31%. These results indicate that DLEPEM effectively mitigates catastrophic forgetting while enhancing incremental learning capability. Full article

Irregular Hangeul typefaces present a challenging computer vision problem because complete font generation must generalize from a small number of reference glyphs while preserving both structural consistency and stylistic fidelity. This study investigates few-shot learning for the restoration and expansion of irregular and historical Hangeul typefaces through three experiments spanning relatively regular woodblock print, irregular contemporary type, and highly irregular royal calligraphy. We benchmark a GAN-based model (DM-Font), a VQGAN-based model (VQ-Font), and a diffusion-based model (Diff-Font) under limited supervision and evaluate them using pixel-level similarity, structural indicator, OCR usability, and expert assessment. DM-Font established a feasible baseline for historical restoration (mean SSIM 0.77), whereas VQ-Font obtained the highest structural similarity for irregular contemporary typeface when paired with a structurally designed 10-character pangram reference set (SSIM 0.97; OCR accuracy 99.5% on the evaluated glyph set). For highly irregular royal calligraphy, the two models performed comparably on global similarity (SSIM 0.78 vs. 0.80) and on expert ratings (4.2 vs. 4.3); VQ-Font showed more stable structure-sensitive indicators, whereas Diff-Font better preserved stylistic nuance. The findings suggest that reference-set composition substantially affects generation quality under fixed-budget few-shot conditions, and that model choice should be matched to source regularity and restoration objectives. Full article

Aircraft skin defect recognition is a safety-critical visual inspection task in which lightweight models must maintain high diagnostic accuracy while suppressing false alarms caused by complex surface textures, illumination variations, and weak defect patterns. This study proposes HQCA-Net, a simulation-based hybrid quantum-classical channel attention network for reliable aircraft skin defect recognition. The core component, termed Residual Quantum Channel Attention (RQCA), embeds a 10-qubit variational quantum circuit into a classical ResNet-18 backbone to perform compact and structured nonlinear feature recalibration, introducing only 30 trainable quantum-gate parameters. The quantum circuit is evaluated using state-vector simulation, and this study focuses on model-level feature recalibration, reliability, and robustness within the evaluated dataset rather than implementation on physical quantum hardware. Experiments on a six-class aircraft skin defect dataset show that HQCA-Net achieves 97.93% classification accuracy and a global false positive rate of 0.49%, outperforming ResNet-18 and classical lightweight attention mechanisms including SE, ECA, and SimAM. Additional analyses using confidence calibration, Grad-CAM visualization, Gaussian noise perturbation, few-shot training, and circuit-depth ablation further indicate that the proposed RQCA module improves feature discrimination and false-alarm suppression under compact parameter constraints. These results suggest that the hybrid quantum-classical attention module can serve as a parameter-efficient nonlinear feature recalibration strategy for reliable visual defect inspection under the tested experimental conditions. Full article

Remote tower operations require air traffic controllers to maintain continuous visual monitoring and integrate information from panoramic displays, radar data, flight strips, and voice communication. Such screen-mediated and sustained surveillance tasks may lead to covert fatigue, which is difficult to capture using a single physiological or behavioral signal. To address this issue, this study proposes a Gated EEG–Eye Fusion Network (GEEF-Net) for window-level fatigue detection in remote tower controllers. EEG and eye-tracking signals were synchronously collected during simulated remote tower tasks and segmented into 5 s windows with a 2 s step. For each window, 53 EEG features and 47 eye-tracking features were extracted to construct a 100-dimensional multimodal representation. GEEF-Net adopts a lightweight modality-gating mechanism to adaptively weight EEG and eye-tracking representations before fatigue classification. Under the main subject-dependent validation setting, GEEF-Net achieved an Accuracy of 0.883, an F1-score of 0.788, and a ROC-AUC of 0.944, outperforming EEG-only, eye-only, and early-fusion baselines in most overall metrics. The gating analysis indicated that eye-tracking features received a higher average weight than EEG features, suggesting the importance of visual behavior in remote tower fatigue detection. Cross-subject validation showed that individual differences remain a major challenge, while few-shot subject-specific calibration improved model adaptation when limited target-subject samples were available. These findings suggest that EEG–eye-tracking fusion with lightweight modality gating is a feasible approach for fatigue detection in simulated remote tower tasks. However, larger datasets and operationally realistic validation considering shift work, circadian effects, and operational pressure are still required before the approach can be considered operationally reliable. Full article

The NMR resonator is the sensor responsible for transmitting RF pulses and receiving detection signals, and its tuning and matching are crucial to acquiring high-sensitivity NMR signals. Automated tuning and matching (ATM) is therefore essential for rapid, accurate, and continuously efficient testing. Existing NMR ATM methods still primarily rely on iterative search strategies, whose dominant cost arises from repeated hardware measurements and waiting periods, often requiring multiple measurement cycles before convergence. The emergence of in situ NMR detection of high-concentration ionic samples has further increased the demand for real-time, rapid ATM with a large dynamic range, posing a major challenge to conventional approaches. This paper proposes a physics-informed few-shot learning method for automatic tuning and matching over wideband and multi-resonance-frequency NMR scenarios. The tuning-and-matching problem is formulated as a structure and frequency-conditioned function regression task, and a conditional neural process (CNP) is introduced to learn cross-task priors and directly predict the states of tunable components from only a small number of real-machine context measurements. A physics regularizer based on the local sensitivity of the input impedance is further designed to impose stronger penalties on errors under high-Q narrowband operating conditions without relying on proprietary analytical circuit models. Simulation studies and real NMR experiments are conducted on multiple circuit topologies and multiple target frequencies using only a small number of NMR samples. The results demonstrate consistent improvements in key metrics, including accuracy of tuning and matching and the number of collected real-machine samples required per task. In particular, with only 100 sampled tuning/matching capacitor points and 20 on-hardware collected samples, the proposed method already delivers satisfactory tuning-and-matching performance. The method achieves an attractive accuracy–cost tradeoff across both cross-topology and cross-frequency scenarios, and shows strong potential for few-shot, rapid, real-time detection. Full article

To address the issues of model overfitting under scarce samples and poor segmentation performance on slender objects in the task of semantic segmentation of remote sensing hyperspectral images, this paper proposes a hyperspectral image semantic segmentation framework that integrates edge awareness and multi-scale feature enhancement under extremely few-shot conditions. This architecture effectively integrates orthogonal-direction convolutions, elongated feature enhancement, multi-scale feature fusion, and deep supervision mechanisms, solving challenges such as difficulty in extracting features of slender objects, model overfitting under few-sample conditions, and insufficient generalization ability. The experimental results on multiple public datasets show that the proposed algorithm achieves excellent segmentation performance with just one small-sized sample per labeled category, surpassing existing popular algorithms and thereby confirming the algorithm’s effectiveness and superiority. On the PaviaU dataset, the overall accuracy (OA) and mean intersection over union (mIoU) improved by approximately 9.7% and 15.5% compared to the second-best model; especially for the segmentation of the key elongated feature ‘road’, the intersection over union reached 94.75%, highlighting the effectiveness of the proposed mechanism. This paper provides a novel and efficient solution for fine interpretation of hyperspectral images under few-sample conditions. Full article

Underwater target detection via side-scan sonar is vital for defense and economy but hindered by sparse targets, high data costs, and feature extraction difficulties due to textureless acoustic data and limited samples. To overcome these limitations, particularly for few-shot, small-object detection, we propose a Spatial Distribution Probability-Guided Detection Framework to aid Unmanned Underwater Vehicles (UUVs) in precise localization and clustering. The framework features a novel module that leverages a pre-trained Vision Foundation Model (DINOv3) to generate spatial distribution probability maps, guiding a Transformer-based network for accurate detection with scarce data. Additionally, it incorporates a Target Position Calculation Module and a DBSCAN-based post-processing module to determine global geographic coordinates and cluster discrete points, respectively. Experiments were conducted on both a Public Mine Detection Dataset and a self-collected dataset containing simulated mines and buoys. Ablation studies and comparison experiments demonstrated that the proposed guidance mechanism significantly improves detection performance. Furthermore, two comb-search missions verified that the system could accurately locate and cluster targets, distinguishing real targets from false detections (noise). These results confirm the framework’s efficacy in enabling high-precision perception and autonomous operations for complex underwater inspection tasks. Full article

In response to the challenges of insufficient multi-scale feature representation and limited model adaptability in soybean leaf disease recognition from field images, a semantic residual Transformer (Sem-ResFormer) model is proposed for soybean leaf disease identification. The proposed model is constructed by integrating multi-scale residual feature extraction, Transformer-based global dependency modeling, and a semantic mapping mechanism, through which effective modeling and semantic representation of multi-scale visual information in lesion regions are achieved. A multimodal large model fine-tuning strategy combined with cross-architecture hyperparameter transfer is employed. The optimal hyperparameter configuration of the Vision Transformer, obtained via Bayesian optimization, is transferred to Qwen2.5-VL, and a progressive fine-tuning strategy is adopted, whereby the adaptability of the model to task-specific data is gradually enhanced. Experiments were conducted on a constructed five-class field-image soybean leaf disease dataset containing 3852 images, with 674 labeled images used in the initial few-shot fine-tuning stage. Under an input resolution of 720 × 720, the proposed method achieved an overall accuracy (OA) of 95.33%, surpassing the OA obtained with the default parameter configuration (93.64%) and the ResNet-50-based transfer method (93.43%). In the initial few-shot stage, the OA was improved from 74.05% under zero-shot conditions to 90.66%. These results demonstrate that the proposed method effectively improves soybean leaf disease recognition accuracy and model adaptability under the constructed field-image dataset with visual variability. Full article

The objective of few-shot segmentation is to segment novel categories given only a few annotated support images. Current FSS methods typically rely on pretrained backbone networks while often overlooking the inherent discrepancy between pretraining tasks and downstream segmentation tasks. This oversight renders the models susceptible to noise interference and hinders rapid generalization to novel categories. To address these limitations, we propose BAFNet, a novel few-shot segmentation algorithm based on two-stage backbone adaptive fine-tuning. Our approach incorporates a Feature Activation Adapter module into the backbone network, which operates through similarity feature enhancement and low-dimensional adaptive learning. Building upon this foundation, we develop an adapter-based fine-tuning strategy for the training phase that enhances the backbone network’s capacity for extracting category-relevant features while optimizing similarity representation of the extracted features. Additionally, we introduce a support set-driven, in-episode, online fine-tuning strategy for the testing phase, which leverages data augmentation to generate pseudo-query sets for supervised fine-tuning optimization. Through comprehensive quantitative and qualitative experiments conducted on PASCAL-5ⁱ, COCO-20ⁱ, and the industrial MT Defect Dataset, our results demonstrate that the proposed BAFNet model achieves state-of-the-art few-shot segmentation performance while utilizing the minimal number of trainable parameters. Our method obtains superior performance for both the mean intersection over union and foreground-background intersection over union evaluation metrics, exhibiting remarkable applicability for both general images in complex scenes and industrial defect segmentation under few-shot conditions. Full article

Plant disease recognition models trained under laboratory conditions often degrade markedly after cross-domain transfer because of the pronounced distribution gap between source and target domains and the scarcity of labeled target-domain samples. To address the transfer task from PlantVillage (PV_100) to PlantDoc, this study develops and evaluates a lightweight cross-domain few-shot plant disease recognition method under a strict PlantVillage-to-PlantDoc protocol. The method integrates EfficientNet-B0 feature extraction, cosine-similarity-based prototypical classification, and training-time target-domain BN adaptation (TBA). During training, unlabeled target-domain images are used only for BN statistical calibration, whereas inference is limited to feature extraction and prototype matching, without gradient updates or iterative optimization. Under a unified experimental protocol, the proposed method achieved cross-split mean accuracies of 42.69 ± 0.62% for one-shot and 54.24 ± 0.72% for five-shot, where ± denotes the standard deviation across three strict data splits; it outperformed ProtoNet by 7.44 and 9.43 percentage points, respectively. Ablation results indicate that TBA is the main source of performance improvement, whereas more complex adaptation strategies do not yield stable additional gains. The core encoder can be executed entirely on the NPU, with an estimated single-sample inference latency as low as 0.658 ms, indicating strong potential for encoder-level mobile deployment. Full article

Deep learning has exhibited potential in land use and land cover (LULC) classification applications. However, the effectiveness of deep learning remains constrained by the availability and quality of annotated training data. The persistent scarcity of labeled samples and spatial heterogeneity of remote sensing imagery hinder the robustness and generalization of trained models. Meta-learning, commonly referred to as “learning to learn”, is a paradigm that trains models over a distribution of tasks to acquire transferable knowledge, enabling rapid adaptation to new tasks with only a few labeled samples. This cross-task learning capability makes meta-learning a promising solution to data scarcity and spatial heterogeneity in the remote sensing context. This paper provides a systematic review of meta-learning applications in LULC classification, identifying a total of 70 relevant studies between 2018 and 2025. Three mainstream meta-learning paradigms (memory-augmented, optimization-based, and metric-based) are reviewed, and the applications are analyzed across four core challenges in LULC remote sensing: label scarcity, cross-region and cross-domain distribution shifts, temporal dynamics modeling, and multimodal data integration. The review reveals that optimization-based and metric-based methods dominate current research, with MAML and its variants being the most widely adopted due to the model-agnostic property, while memory-augmented methods remain underexplored. A consistent finding is that meta-learning outperforms conventional pre-training followed by fine-tuning under significant domain shifts across multiple data modalities. Current limitations, including computational overhead, episodic training constraints, and the lack of standardized evaluation protocols, are discussed. Future directions in cross-domain generalization, integration with foundation models, novel architectures, and standardized benchmarks are identified. Full article

Accurate mapping of crop planting structures is critical for precision agriculture, yet it remains challenging in rugged terrain with fragmented fields, frequent cloud contamination, and limited high-quality training samples. This study evaluates an integrated framework combining recursive feature elimination (RFE) and the pretrained Tabular Prior-Data Fitted Network (TabPFN) for small-sample crop classification using Sentinel-2 time-series data in Yuxi City, located on the western margin of the Yunnan–Guizhou Plateau. A multidimensional feature set integrating spectral and temporal vegetation indices and textural and geospatial information was constructed and optimized via RFE. The TabPFN model achieved an overall accuracy (OA) of 96.27%, a kappa coefficient of 0.9558, and a macro-F1 score of 0.956 in the main validation. In repeated small-sample experiments, TabPFN maintained a mean OA of 90.60% at a 30% training-sample ratio and 82.89% at a 10% ratio. RF-guided feature ranking and ablation analyses suggested that temporal vegetation indices were important predictors, followed by early-season spectral characteristics, textural features, and supplementary geospatial information. Overall, these findings indicate that RFE-TabPFN is a feasible option for 10 m crop mapping in Yuxi under limited training samples, while its broader applicability still requires further testing across additional years, regions, and cropping systems. Full article