Search Results (185)

Hyperspectral image change detection (HSI-CD) provides substantial value in environmental monitoring, urban planning and other fields. In recent years, deep-learning based HSI-CD methods have made remarkable progress due to their powerful nonlinear feature learning capabilities, yet they face several challenges: mixed-pixel phenomenon affecting pixel-level detection accuracy; heterogeneous spatial scales of change targets where coarse-grained features fail to preserve fine-grained details; and dependence on high-quality labels. To address these challenges, this paper introduces WSCDNet, a weakly supervised HSI-CD network employing coarse-to-fine feature learning, with key innovations including: (1) A dual-branch detection framework integrating binary and multiclass change detection at the sub-pixel level that enhances collaborative optimization through a cross-feature coupling module; (2) introduction of multi-granularity aggregation and difference feature enhancement module for detecting easily confused regions, which effectively improves the model’s detection accuracy; and (3) proposal of a weakly supervised learning strategy, reducing model sensitivity to noisy pseudo-labels through decision-level consistency measurement and sample filtering mechanisms. Experimental results demonstrate that WSCDNet effectively enhances the accuracy and robustness of HSI-CD tasks, exhibiting superior performance under complex scenarios and weakly supervised conditions. Full article

Semi-supervised object detection (SSOD) in remote sensing is challenged by the accumulation of pseudo-label errors in complex scenes with dense objects and high intra-class variability. While teacher–student frameworks enable learning from unlabeled data, erroneous pseudo-labels such as false positives and missed detections can be reinforced over time, which degrades model performance. To address this issue, we propose the Error-Mitigation Teacher (EMT), a unified framework designed to suppress error propagation during SSOD training. EMT consists of three lightweight modules. First, the Adaptive Pseudo-Label Filtering (APLF) module removes noisy pseudo boxes via a second-stage RCNN and adjusts class-specific thresholds through dynamic confidence filtering. Second, the Confidence-Based Loss Reweighting (CBLR) module reweights training loss by evaluating the teacher model’s ability to reconstruct its own pseudo-labels, using the resulting loss as an indicator of label reliability. Third, the Enhanced Supervised Learning (ESL) module improves class-level balance by adjusting supervised loss weights according to pseudo-label statistics. EMT demonstrates consistent performance gains over representative state-of-the-art SSOD methods on DOTA, DIOR, and SSDD datasets. Notably, EMT achieves a 2.9% absolute mAP₅₀ improvement on DIOR using only 10% of labeled data, without incurring additional inference cost. These results highlight EMT’s effectiveness in improving SSOD for remote sensing. Full article

The transformer-based deep learning approach represents the current state-of-the-art in machine translation (MT) research. Large-scale pretrained transformer models produce state-of-the-art performance across a wide range of MT tasks for many languages. However, such deep neural network (NN) models are often data-, compute-, space-, power-, and energy-hungry, typically requiring powerful GPUs or large-scale clusters to train and deploy. As a result, they are often regarded as “non-green” and “unsustainable” technologies. Distilling knowledge from large deep NN models (teachers) to smaller NN models (students) is a widely adopted sustainable development approach in MT as well as in broader areas of natural language processing (NLP), including speech, and image processing. However, distilling large pretrained models presents several challenges. First, increased training time and cost that scales with the volume of data used for training a student model. This could pose a challenge for translation service providers (TSPs), as they may have limited budgets for training. Moreover, CO₂ emissions generated during model training are typically proportional to the amount of data used, contributing to environmental harm. Second, when querying teacher models, including encoder–decoder models such as NLLB, the translations they produce for low-resource languages may be noisy or of low quality. This can undermine sequence-level knowledge distillation (SKD), as student models may inherit and reinforce errors from inaccurate labels. In this study, the teacher model’s confidence estimation is employed to filter those instances from the distilled training data for which the teacher exhibits low confidence. We tested our methods on a low-resource Urdu-to-English translation task operating within a constrained training budget in an industrial translation setting. Our findings show that confidence estimation-based filtering can significantly reduce the cost and CO₂ emissions associated with training a student model without drop in translation quality, making it a practical and environmentally sustainable solution for the TSPs. Full article

In response to the limitations of traditional image processing algorithms, such as high noise sensitivity and threshold dependency in bridge crack detection, and the extensive labeled data requirements of deep learning methods, this study proposes a novel crack segmentation algorithm based on fuzzy C-means (FCM) clustering and multi-feature fusion. A three-dimensional feature space is constructed using B-channel pixels and fuzzy clustering with $c$ = 3, justified by the distinct distribution patterns of these three regions in the image, enabling effective preliminary segmentation. To enhance accuracy, connected domain labeling combined with a circularity threshold is introduced to differentiate linear cracks from granular noise. Furthermore, a 5 $\times$ 5 neighborhood search strategy, based on crack pixel amplitude, is designed to restore the continuity of fragmented cracks. Experimental results on the Concrete Crack and SDNET2018 datasets demonstrate that the proposed algorithm achieves an accuracy of 0.885 and a recall rate of 0.891, outperforming DeepLabv3+ by 4.2%. Notably, with a processing time of only 0.8 s per image, the algorithm balances high accuracy with real-time efficiency, effectively addressing challenges, such as missed fine cracks and misjudged broken cracks in noisy environments by integrating geometric features and pixel distribution characteristics. This study provides an efficient unsupervised solution for bridge damage detection. Full article

Class-incremental learning (CIL) in deep neural networks is affected by catastrophic forgetting (CF), where acquiring knowledge of new classes leads to the significant degradation of previously learned representations. This challenge is particularly severe in medical image analysis, where costly, expertise-dependent annotations frequently contain pervasive and hard-to-detect noisy labels that substantially compromise model performance. While existing approaches have predominantly addressed CF and noisy labels as separate problems, their combined effects remain largely unexplored. To address this critical gap, this paper presents a dual-stage clean-sample selection method for Incremental Noisy Label Learning (DSCNL). Our approach comprises two key components: (1) a dual-stage clean-sample selection module that identifies and leverages high-confidence samples to guide the learning of reliable representations while mitigating noise propagation during training, and (2) an experience soft-replay strategy for memory rehearsal to improve the model’s robustness and generalization in the presence of historical noisy labels. This integrated framework effectively suppresses the adverse influence of noisy labels while simultaneously alleviating catastrophic forgetting. Extensive evaluations on public medical image datasets demonstrate that DSCNL consistently outperforms state-of-the-art CIL methods across diverse classification tasks. The proposed method boosts the average accuracy by 55% and 31% compared with baseline methods on datasets with different noise levels, and achieves an average noise reduction rate of 73% under original noise conditions, highlighting its effectiveness and applicability in real-world medical imaging scenarios. Full article

Label Distribution Learning (LDL) has emerged as a powerful paradigm for addressing label ambiguity, offering a more nuanced quantification of the instance–label relationship compared to traditional single-label and multi-label learning approaches. This paper focuses on the challenge of noisy label distributions, which is ubiquitous in real-world applications due to the annotator subjectivity, algorithmic biases, and experimental errors. Existing related LDL algorithms often assume a linear combination of true and random label distributions when modeling the noisy label distributions, an oversimplification that fails to capture the practical generation processes of noisy label distributions. Therefore, this paper introduces an assumption that the noise in label distributions primarily arises from the semantic confusion between labels and proposes a novel generative label distribution learning algorithm to model the confusion-based generation process of both the feature data and the noisy label distribution data. The proposed model is inferred using variational methods and its effectiveness is demonstrated through extensive experiments across various real-world datasets, showcasing its superiority in handling noisy label distributions. Full article

In this paper, we propose TransConvNet, a hybrid model combining Convolutional Neural Networks (CNNs), self-attention mechanisms, and transfer learning for wireless signal recognition under challenging conditions. The model effectively addresses challenges such as low signal-to-noise ratio (SNR), low sampling rates, and limited labeled data. The CNN module extracts local features and suppresses noise, while the self-attention mechanism within the Transformer encoder captures long-range dependencies in the signal. To enhance performance with limited data, we incorporate transfer learning by leveraging pre-trained models, ensuring faster convergence and improved generalization. Extensive experiments were conducted on a six-class wireless signal dataset, downsampled to 1 MSPS to simulate real-world constraints. The proposed TransConvNet achieved 92.1% accuracy, outperforming baseline models such as LSTM, CNN, and RNN across multiple evaluation metrics, including RMSE and $R^2$. The model demonstrated strong robustness under varying SNR conditions and exhibited superior discriminative ability, as confirmed by Precision–Recall and ROC curves. These results validate the effectiveness and robustness of the TransConvNet model for wireless signal recognition, particularly in resource-constrained and noisy environments. Full article

State estimation in distribution power systems is increasingly challenged by the proliferation of distributed energy resources (DERs), bidirectional power flows, and the growing complexity of unbalanced network topologies. Physics-Informed Neural Networks (PINNs) offer a compelling solution by integrating machine learning with the physical laws that govern power system behavior. This paper introduces a PINN-based framework for state estimation in unbalanced distribution systems, leveraging available data and embedded physical knowledge to improve accuracy, computational efficiency, and robustness across diverse operating scenarios. The proposed method is evaluated on four IEEE test feeders—IEEE 13, 34, 37, and 123—using synthetic datasets generated via OpenDSS to emulate realistic operating scenarios, and demonstrates significant improvements over baseline models. Notably, the PINN achieves up to a 97% reduction in current estimation errors while maintaining high voltage prediction accuracy. Extensive simulations further assess model performance under noisy inputs and partial observability, where the PINN consistently outperforms conventional data-driven approaches. These results highlight the method’s ability to generalize under uncertainty, accelerate convergence, and preserve physical consistency in simulated real-world conditions without requiring large volumes of labeled training data. Full article

The separation of respiratory and cardiac sounds is a significant challenge in biomedical signal processing due to their overlapping frequency and time characteristics. Traditional methods struggle with accurate extraction in noisy or diverse clinical environments. This study explores the application of machine learning, particularly convolutional neural networks (CNNs), to overcome these obstacles. Advanced machine learning models, denoising algorithms, and domain adaptation strategies address challenges such as frequency overlap, external noise, and limited labeled datasets. This study presents a robust methodology for detecting heart and lung diseases from audio signals using advanced preprocessing, feature extraction, and deep learning models. The approach integrates adaptive filtering and bandpass filtering as denoising techniques and variational autoencoders (VAEs) for feature extraction. The extracted features are input into a CNN, which classifies audio signals into different heart and lung conditions. The results highlight the potential of this combined approach for early and accurate disease detection, contributing to the development of reliable diagnostic tools for healthcare. Full article

Small-scale object detection remains a significant challenge in semi-supervised object detection (SSOD), particularly in panoramic dental X-rays. Due to the small lesion size, low contrast, and complex anatomical background, conventional teacher models often fail to extract accurate lesion features, leading to noisy pseudo labels and suboptimal detection performance. Additionally, most existing SSOD methods rely on high-confidence thresholds to select pseudo labels, which may mistakenly discard valuable predictions with low scores but accurate localization—especially for small-scale targets. To address these challenges, we propose Entropy Teacher, a novel SSOD framework specifically designed for small-scale dental disease detection. Our method introduces an Entropy-Guided Feature Pyramid that integrates entropy-guided representations to enhance fine-grained structural learning. Moreover, we develop a low-confidence pseudo-label mining (LCPLM) strategy with a class-adaptive thresholding mechanism to effectively recover high-quality pseudo labels below conventional confidence thresholds. Extensive experiments on the Dental Disease Dataset and ChestX-Det demonstrate that Entropy Teacher achieves state-of-the-art performance, surpassing the baseline Unbiased Teacher by +3.8 $A{P}_{50}$ and +4.5 $A{P}_S$. These results confirm the effectiveness of entropy-guided representations and low-confidence mining in improving small-scale lesion detection under limited supervision. Full article

Remote photoplethysmography (rPPG) enables non-contact physiological measurement for emotion recognition, yet the temporally sparse nature of emotional cardiovascular responses, intrinsic measurement noise, weak session-level labels, and subtle correlates of valence pose critical challenges. To address these issues, we propose a physiologically inspired deep learning framework comprising a Multi-scale Temporal Dynamics Encoder (MTDE) to capture autonomic nervous system dynamics across multiple timescales, an adaptive sparse α-Entmax attention mechanism to identify salient emotional segments amidst noisy signals, Gated Temporal Pooling for the robust aggregation of emotional features, and a structured three-phase curriculum learning strategy to systematically handle temporal sparsity, weak labels, and noise. Evaluated on the MAHNOB-HCI dataset (27 subjects and 527 sessions with a subject-mixed split), our temporal-only model achieved competitive performance in arousal recognition (66.04% accuracy; 61.97% weighted F1-score), surpassing prior CNN-LSTM baselines. However, lower performance in valence (62.26% accuracy) revealed inherent physiological limitations regarding a unimodal temporal cardiovascular analysis. These findings establish clear benchmarks for temporal-only rPPG emotion recognition and underscore the necessity of incorporating spatial or multimodal information to effectively capture nuanced emotional dimensions such as valence, guiding future research directions in affective computing. Full article

Currently, samples are a critical driving force in the application of deep learning. However, the use of samples encounters problems, such as an inconsistent annotation quality, mismatches between images and labels, and a lack of fine-grained labels. Refining sample labels is essential for training a sophisticated model. Refining sample labels through manual verification is labor-intensive, especially for training large models. Additionally, existing label refinement methods based on deep neural networks (DNNs) typically rely on image features to directly predict segmentation results, often overlooking the potential information embedded in existing noisy labels. To address these challenges and shortcomings, this study proposes a novel remote sensing sample label refinement (LR) network, named the identify–update–refine network (IUR-Net). IUR-Net leverages newly acquired remote sensing images and their corresponding noisy labels to automatically identify erroneous regions, update them with more accurate information, and refine the results to improve label quality. A multi-scale, error-aware localization module (Ms-EALM) is designed to capture label–image inconsistencies, enabling the more accurate localization of erroneous label regions. To evaluate the proposed framework, we first constructed and publicly released two benchmark datasets for the label refinement task: WHU-LR and EVLAB-LR. The experimental results on these datasets demonstrate that the labels refined by IUR-Net not only outperform the baseline model in both IoU and F1 scores, but also effectively identify errors in noisy annotations. Full article

Deep learning-based models have achieved considerable success in partial discharge (PD) fault diagnosis for power systems, enhancing grid asset safety and improving reliability. However, traditional approaches often rely on centralized training, which demands significant resources and fails to account for the impact of noisy operating conditions on Intelligent Electronic Devices (IEDs). In a gas-insulated switchgear (GIS), PD measurement data collected in noisy environments exhibit diverse feature distributions and a wide range of class representations, posing significant challenges for trained models under complex conditions. To address these challenges, we propose a Self-Supervised Asynchronous Federated Learning (SSAFL) approach for PD diagnosis in noisy IED environments. The proposed technique integrates asynchronous federated learning with self-supervised learning, enabling IEDs to learn robust pattern representations while preserving local data privacy and mitigating the effects of resource heterogeneity among IEDs. Experimental results demonstrate that the proposed SSAFL framework achieves overall accuracies of 98% and 95% on the training and testing datasets, respectively. Additionally, for the floating class in IED 1, SSAFL improves the F1-score by 5% compared to Self-Supervised Federated Learning (SSFL). These results indicate that the proposed SSAFL method offers greater adaptability to real-world scenarios. In particular, it effectively addresses the scarcity of labeled data, ensures data privacy, and efficiently utilizes heterogeneous local resources. Full article

Accurate fish species identification is crucial for marine biodiversity conservation, environmental monitoring, and sustainable fishery management, particularly as marine ecosystems face increasing pressures from human activities and climate change. Traditional morphological identification methods are inherently labor-intensive and resource-demanding, while contemporary automated approaches, particularly deep learning models, often suffer from significant computational overhead and struggle with the pervasive issue of class imbalance inherent in ecological datasets. Addressing these limitations, this research introduces a novel computationally parsimonious fish classification framework leveraging the hybrid Mobile–Former neural network architecture. This architecture strategically combines the local feature extraction strengths of convolutional layers with the global context modeling capabilities of transformers, optimized for efficiency. To specifically mitigate the detrimental effects of the skewed data distributions frequently observed in real-world fish surveys, the framework incorporates a sophisticated robust asymmetric loss function designed to enhance model focus on under-represented categories and improve resilience against noisy labels. The proposed system was rigorously evaluated using the comprehensive FishNet dataset, comprising 74,935 images distributed across a detailed taxonomic hierarchy including eight classes, seventy-two orders, and three-hundred-forty-eight families, reflecting realistic ecological diversity. Our model demonstrates superior classification accuracy, achieving 93.97 percent at the class level, 88.28 percent at the order level, and 84.02 percent at the family level. Crucially, these high accuracies are attained with remarkable computational efficiency, requiring merely 508 million floating-point operations, significantly outperforming comparable state-of-the-art models in balancing performance and resource utilization. This advancement provides a streamlined, effective, and resource-conscious methodology for automated fish species identification, thereby strengthening ecological monitoring capabilities and contributing significantly to the informed conservation and management of vital marine ecosystems. Full article

Convolutional neural networks (CNNs) and their application in animal husbandry have in-depth mathematical expressions, which usually revolve around how well they map input data such as images or video frames of animals to meaningful outputs like health status, behavior class, and identification. Likewise, computer vision and deep learning models are driven by CNNs to act intelligently in improving productivity and animal management for sustainable animal husbandry. In animal husbandry, CNNs play a vital role in the management and monitoring of livestock’s health and productivity due to their high-performance accuracy in analyzing images and videos. Monitoring animals’ health is important for their welfare, food abundance, safety, and economic productivity. This paper aims to comprehensively review recent advancements and applications of relevant models that are based on CNNs for livestock health monitoring, covering the detection of their various diseases and classification of their behavior, for overall management gain. We selected relevant articles with various experimental results addressing animal detection, localization, tracking, and behavioral monitoring, validating the high-performance accuracy and efficiency of CNNs. Prominent anchor-based object detection models such as R-CNN (series), YOLO (series) and SSD (series), and anchor-free object detection models such as key-point based and anchor-point based are often used, demonstrating great versatility and robustness across various tasks. From the analysis, it is evident that more significant research contributions to animal husbandry have been made by CNNs. Limited labeled data, variation in data, low-quality or noisy images, complex backgrounds, computational demand, species-specific models, high implementation cost, scalability, modeling complex behaviors, and compatibility with current farm management systems are good examples of several notable challenges when applying CNNs in animal husbandry. By continued research efforts, these challenges can be addressed for the actualization of sustainable animal husbandry. Full article