Search Results (602)

Unsupervised domain adaptation (UDA) remains challenged by an unstable target structure, pseudo-label noise, and heterogeneous transfer difficulty across domains. To address these issues, we propose Progressive Multi-View Graph Projection (PMGP), a two-stage framework that first learns transferable representations via source supervision, domain-adversarial training, and teacher–student consistency and then performs latent-space refinement through multi-view graph construction and projection learning. Specifically, three perturbation-induced views are considered for each sample: the original view, an input-space patch-masked view, and a representation-space feature-dimension masked view. After joint preprocessing with PCA and L2 normalization, PMGP constructs per-view affinity graphs by combining geometric neighborhood relations with pseudo-supervised semantic relations, and applies locality-preserving projection to learn a structure-aware shared subspace. In this subspace, target pseudo-labels are iteratively refined using source prototypes, target class centers, and progressive confidence filtering. Experiments on Office-Home, ImageCLEF-DA, and VisDA-2017 show that PMGP achieves competitive performance and stable behavior across different benchmark settings and backbone architectures. These results indicate that multi-view graph refinement provides an effective and interpretable way to improve target structure estimation and reduce pseudo-label error accumulation in UDA. Full article

Adverse weather and low illumination remain major challenges for autonomous driving perception, where semantic segmentation must stay reliable despite severe appearance degradation. In unsupervised domain adaptation without target annotations, self-training is widely used, but it is often limited by the inconsistent quality of teacher-generated pseudo labels across samples, regions, and training stages. This paper presents RaDA, a reliability-aware self-training framework that regulates pseudo supervision at three levels. First, a progressive exposure strategy determines which target images are admitted for training. Second, spatial reliability weighting suppresses gradients from degraded regions while retaining informative supervision. Third, adaptive teacher update scheduling stabilizes pseudo label generation over time. Experiments on real-world adverse driving benchmarks show that RaDA improves robustness, training stability, and cross-dataset generalization compared with strong baselines. Compared with the previous state-of-the-art method MIC, RaDA achieves mIoU gains of 10.6 percentage points on Foggy Zurich and 8.8 percentage points on the Foggy Driving benchmark. These results indicate that explicit reliability regulation can strengthen self-training domain adaptation for semantic segmentation in autonomous driving under challenging environmental conditions. Full article

Deep neural networks suffer performance drops when source and target domains differ in distribution, motivating research into domain adaptation (DA). Traditional DA approaches presume source samples come from a single domain and can be available during adaptation. Nevertheless, in real-world applications, multiple source domains often exist, and source samples may be inaccessible owing to privacy and storage limitations. In response to the challenges of multi-source and source-free, multi-source-free domain adaptation (MSFDA) is proposed, which captures transferable information from a set of pre-trained source models to boost performance of the model on target domain. Most MSFDA methods meet these challenges by utilizing pseudo-labeling. However, pseudo-labels generated by distinct source models may contain noise and even be contradictory, which weakens their efficacy in facilitating source models adapting to the target domain. Moreover, these methods do not consider class imbalance, which would lead to biased predictions for minority classes, and undermine adaptation. Therefore, we propose a novel MSFDA method which extends adversarial learning to a multi-source-free setting. This method presents proxy multi-source domain adversarial learning, which aligns target features extracted by different source models in an adversarial manner, enhancing the capability of source models to extract domain-invariant features and potentially obtain high-quality pseudo-labels. Moreover, a nuclear-norm maximization regularization is employed to constrain prediction matrices, which can reduce the prediction uncertainty and enhance the discriminability of the model, while mitigating the prediction bias and promoting the prediction accuracy for minority classes. Finally, comprehensive evaluations on four benchmark datasets prove the validity of the proposed method. Full article

Cross-view image geolocalization aims to achieve accurate localization of geo-tagged images without geo-tagging by matching ground-view images with satellite images. However, there are huge imaging differences between ground and satellite viewpoints, and existing methods usually rely on a large number of accurately labeled cross-view image pairs. Therefore, to address issues such as significant perspective differences, high annotation costs, and low utilization of unpaired data, this paper proposes a cross-view generation model that integrates multi-scale contrastive learning and dynamic optimization, designs a multi-scale contrast loss function to strengthen the semantic consistency between the generated images and the target domain, adaptively balances the quality and quantity of pseudo-labels according to a dynamic threshold screening mechanism, and introduces a hard-sample triplet loss to enhance the model discriminative ability. Ablation experiments on the CVUSA and CVACT datasets show that the BEV-CycleGAN+CL (Bird’s-Eye View Cycle-Consistent Generative Adversarial Network with Contrastive Learning) model proposed in this paper significantly outperforms the comparative models in PSNR, SSIM, and RMSE metrics. Specifically, on the CVACT dataset, compared with the BEV-CycleGAN, BEV, and CycleGAN baselines, PSNR increased by 2.83%, 16.02%, and 42.30%, SSIM increased by 6.12%, 8.00%, and 18.48%, and RMSE decreased by 9.28%, 15.51%, and 25.35%, respectively. Similar advantages are observed on the CVUSA dataset. Compared with current state-of-the-art models, the dynamic threshold pseudo-label localization method in this paper demonstrates overall superiority in recall metrics such as R@1, R@5, R@10, and R@1%, for example achieving an R@1 of 98.94% on CVUSA, outperforming the best comparative model, Sample4G^†, which reached 98.68%. This study provides innovative methodological support for disaster emergency response, high-precision map construction for autonomous driving, military reconnaissance, and other applications. Full article

Unsupervised change detection (UCD) from heterogeneous bitemporal optical–SAR imagery is challenging due to modality discrepancy, speckle/illumination variations, and the absence of change annotations. We propose MV-S2CD, a vision foundation model (VFM)-based framework that learns a modality-bridged latent space and produces dense change maps in a fully unsupervised manner. To robustly adapt pretrained VFM priors to heterogeneous inputs with minimal task-specific parameters, MV-S2CD incorporates lightweight modality-specific adapters and parameter-efficient low-rank adaptation (LoRA) in high-level layers. A shared projector embeds the two observations into a common geometry, enabling consistent cross-modal comparison and reducing sensor-induced domain shift. Building on the bridged representation, we design a dual-branch change reasoning module that decouples structure-sensitive cues from semantic-consistency cues: a structure pathway preserves fine boundaries and local variations, while a semantic-consistency pathway employs reliability gating and multi-scale context aggregation to suppress pseudo-changes caused by modality-specific nuisances and residual misregistration. For label-free optimization, we develop a difference-centric self-supervision scheme with two perturbation views and reliability-guided pseudo-partitioning, jointly enforcing pseudo-unchanged invariance, pseudo-changed/unchanged separability, and sparsity and edge-preserving regularization. Experiments on three heterogeneous optical–SAR benchmarks demonstrate that MV-S2CD consistently improves the Precision–Recall trade-off and achieves state-of-the-art performance among unsupervised baselines, while remaining backbone-flexible and efficient. Full article

Background: Automated monitoring of crop health and the precise detection of abiotic stress, such as herbicide damage, are demanding challenges for modern agriculture. Abiotic stresses are a demanding challenge for modern agriculture, responsible for up to 82% of yield losses in major food crops. To address this, researchers are increasingly leveraging artificial intelligence (AI) to automate the detection and management of these stressors. Methods: In particular, this paper presents an instance segmentation framework to precisely detect interveinal chlorosis and leaf curling on potato leaves, two common symptoms of herbicide damage and soft wind. Within the context of precision agriculture and the need to address the inherent ambiguity in manual leaf assessment, this study employs a partial label learning approach to refine the dataset. This method utilizes an EfficientNet-b1 model to classify ambiguous samples, generating high-confidence pseudo-labels for instances that are difficult to categorize visually. The core of the proposed framework is a Mask2Former model, which is first fine-tuned on general potato leaf dataset to enhance its segmentation capabilities and then transferred on the refined, pseudo-labeled dataset. Results & Conclusions: This two-stage approach yields a highly accurate segmentation tool, achieving 89% mAP₅₀ and a pseudo-label classification accuracy of 95%, designed for integration into smart agriculture systems like ground level robotics or unmanned aerial vehicles for real-time, automated crop monitoring. Full article

Predicting the future impact of technical economic documents at release time is challenging due to delayed supervision signals, long-tailed label distributions, and time- and domain-dependent shifts in language and topics. Moreover, similarity graphs derived from text embeddings can be noisy due to boilerplate and evolve under temporal drift, making robustness and leakage-free evaluation essential. We formulate grant-time patent impact prediction as a node classification and within-domain ranking problem on a large-scale semantic similarity document graph built from patent text embeddings, avoiding any future citation leakage. The document graph is constructed via ANN Top-K retrieval and similarity thresholding, enabling scalable and reproducible sparsification on hundreds of thousands of nodes. We propose GraphGPT-Patent, which adapts a reversible graph-to-sequence foundation backbone to local subgraphs extracted from the similarity network. The model incorporates time- and domain-conditioned edge reliability to suppress drift-induced and template-driven pseudo-similarity, and optimizes a joint objective coupling high-impact classification with ranking consistency within comparable groups. Experiments on USPTO granted patents (2000–2022) across three high-volume CPC domains and three evaluation horizons show consistent gains over text-only and GNN baselines, achieving up to 0.94 recall for the positive class and improved macro-average recall across nine settings. Temporal shift analyses further quantify the effect of training-data freshness, while explanation subgraphs provide auditable structural evidence of model decisions. The proposed framework offers an effective graph-based learning pipeline for scalable impact prediction and downstream triage under strict information constraints. Full article

In-depth research on marine biodiversity is essential for understanding and protecting marine ecosystems, where semantic segmentation of marine species plays a crucial role. However, segmenting microscopic zooplankton images remains challenging due to highly variable morphologies, complex boundaries, and the scarcity of high-quality pixel-level annotations that require expert knowledge. Existing semi-supervised methods often rely on single-model perspectives, producing unreliable pseudo-labels and limiting performance in such complex scenarios. To address these challenges, this paper proposes a consistency-driven dual-teacher framework tailored for zooplankton segmentation. Two heterogeneous teacher networks are employed: one captures global morphological features, while the other focuses on local fine-grained details, providing complementary and diverse supervision and alleviating overfitting under limited annotations. In addition, a dynamic fusion-based pseudo-label filtering strategy is introduced to adaptively integrate hard and soft labels by jointly considering prediction consistency and confidence scores, thereby enhancing supervision flexibility. Extensive experiments on the Zooplankton-21 Microscopic Segmentation Dataset (ZMS-21), a self-constructed microscopic zooplankton dataset demonstrate that the proposed method consistently outperforms existing semi-supervised segmentation approaches under various annotation ratios, achieving mIoU scores of 64.80%, 69.58%, 70.32%, and 73.92% with 1/16, 1/8, 1/4, and 1/2 labeled data, respectively. Full article

Small pest detection in ultra-high-resolution forestry images is challenging due to extreme scale variation, complex backgrounds, and limited annotated data. To address these issues, we propose SSFPDet (Semi-Supervised Forest Pest Detector), a semi-supervised object detection framework designed for low-annotation settings. Built upon the Soft Teacher paradigm, SSFPDet integrates a YOLO-T-based overlapping slicing strategy, a Top-K pseudo-label selection mechanism, and a Kullback–Leibler (KL) divergence-based distribution alignment constraint. The slicing strategy enhances small-object representation without modifying the detector backbone, while the Top-K and KL modules improve pseudo-label reliability and semantic consistency during training. Under the 20% labeled setting, SSFPDet achieves an mAP@0.5:0.95 of 46.6, outperforming the baseline by 0.7 points. Notably, small-object detection performance (AP_S) improves by 6.6 percentage points. Ablation studies confirm the complementary contributions of spatial slicing and semantic alignment. Overall, SSFPDet provides a practical and scalable solution for high-resolution forestry pest monitoring under limited supervision. Full article

Unsupervised visible–infrared person re-identification (USL-VI-ReID) plays a pivotal role in cross-modal computer vision applications for intelligent surveillance and public safety. However, the task remains hampered by large modality gaps and limited granularity in feature representations. In particular, channel augmentation (CA) is typically used only for data augmentation, and its potential as an independent input modality remains unexplored. To address these shortcomings, we present a multimodule collaborative USL-VI-ReID framework that explicitly treats CA as a separate input modality. The framework combines four complementary modules. The Person-ReID Adaptive Convolutional Block Attention Module (PA-CBAM) module extracts discriminative features using a two-level attention mechanism that refines salient spatial and channel cues. The Varied Regional Alignment (VRA) module performs cross-modal regional alignment and leverages the Multimodal Assisted Adversarial Learning (MAAL) to reinforce region-level correspondence. The Varied Regional Neighbor Learning (VRNL) implements reliable neighborhood learning via multi-region association to stabilize pseudo-labels and capture local structure. Finally, the Uniform Merging (UM) module merges split clusters through alternating contrastive learning to improve cluster consistency. We evaluate the proposed method on SYSU-MM01 and RegDB. On RegDB’s visible-to-infrared setting, the approach achieves Rank-1 = 93.34%, mean Average Precision (mAP) = 87.55%, and mean Inverse Negative Penalty (mINP) = 76.08%. These results indicate that our method effectively reduces modal discrepancies and increases feature discriminability. It outperforms most existing unsupervised baselines and several supervised approaches, thereby advancing the practical applicability of USL-VI-ReID. Full article

Remote sensing scene classification is one of the crucial techniques for high-resolution remote sensing image interpretation and has received widespread attention in recent years. However, acquiring high-quality labeled data is both costly and time-consuming, making unsupervised domain adaptation (UDA) an important research focus in scene classification. Existing UDA methods focus primarily on aligning the overall feature distributions across domains but neglect class feature alignment, resulting in the loss of critical class information. To address this issue, a cross-layer feature fusion and attention-based class feature alignment network (CFACA-NET) is proposed for unsupervised cross-domain remote sensing scene classification. Specifically, a multi-layer feature extraction module (MFEM) consisting of a cross-layer feature fusion module (CFFM), a multi-scale dynamic attention module (MSDAM), and a fused feature optimization module (FFOM) is designed to enhance the representation ability of scene features. A high-confidence sample selection module is further introduced, which utilizes evidence theory and information entropy to obtain reliable pseudo-labels. Finally, a class feature alignment module is proposed, incorporating a two-stage training strategy to achieve effective class feature alignment. Experimental results on three remote sensing scene classification datasets demonstrate that CFACA-NET outperforms existing state-of-the-art methods in cross-domain classification performance, effectively enhancing cross-domain adaptation capability. Full article

Background: Accurate in vivo brain tumor detection using hyperspectral imaging (HSI), a non-invasive technique that captures spectral information beyond the visible range, is challenging due to the complexity of biological tissues and the difficulty in distinguishing malignant from healthy areas. Conventional neural-network-based methods often misclassify tumor tissue as blood vessels, largely due to high vascularization and the scarcity of annotated data. Method: To address this issue, this work proposes an underexplored approach that decomposes the problem into two tasks: (1) segmentation of the brain cortical surface and its blood vessels, and (2) segmentation of biological tissues within the segmented craniotomy site. The cortical segmentation task is addressed independently of the segmentation model used in the second stage. To achieve this, a set of pseudo-labels is generated from RGB and HSI captures acquired during in vivo brain surgeries. These pseudo-labels support a multimodal training strategy that leverages both imaging domains, yielding a model capable of segmenting the craniotomy site and the blood vessels contained in it. The model is further refined on HSI using weakly supervised fine-tuning with sparse ground truth annotations. Results: The final segmentation map combines cortical and tissue segmentation outputs, considering only cortex pixels not overlapped by vessels as potential tumor regions. This simplifies the HSI tissue segmentation task, reframing it as a binary segmentation of healthy vs. other tissues, while still enabling a comprehensive multiclass output. Conclusions: The proposed method achieves up to a 15.48% increase in F1 score for the tumor class, while segmenting the brain cortex with a mean Dice similarity coefficient (DSC) of 92.08% and accurately detecting 95.42% of labeled blood vessel samples in the HSI dataset. Full article

The automated management of invasive weeds is critical for sustainable agriculture, yet the performance of deep learning models in real-world fields is often compromised by two factors: challenging environmental conditions and the high cost of data annotation. This study tackles both issues through a diagnostic-driven, semi-supervised framework. Using a unique dataset of approximately 975 labelled and 10,000 unlabelled images of Guinea Grass in sugarcane, we first establish strong supervised baselines for classification (ResNet) and detection (YOLO, RF-DETR), achieving F1 scores up to 0.90 and mAP50 scores exceeding 0.82. Crucially, this foundational analysis, aided by interpretability tools, uncovered a pervasive “shadow bias,” where models learned to misidentify shadows as vegetation. This diagnostic insight motivated our primary contribution: a semi-supervised pipeline that leverages unlabelled data to enhance model robustness. By training models on a more diverse set of visual information through pseudo-labelling, this framework not only helps mitigate the shadow bias but also provides a tangible boost in recall, a critical metric for minimising weed escapes in automated spraying systems. To validate our methodology, we demonstrate its effectiveness in a low-data regime on a public crop–weed benchmark. Our work provides a clear and field-tested framework for developing, diagnosing, and improving robust computer vision systems for the complex realities of precision agriculture. Full article

Mangrove extent has increased in many regions under strengthened conservation policies and large-scale restoration programs. Nevertheless, mangrove ecosystems continue to face multiple pressures, including limited total area, habitat degradation, biodiversity decline, and biological invasion, and localized deterioration in ecosystem structure and function has been increasingly reported. Despite extensive mapping efforts, the spatiotemporal dynamics of mangrove degradation—particularly in tidally influenced environments—remain insufficiently understood. Focusing on the Nanliu River Delta, the largest deltaic mangrove system in the Northern Beibu Gulf of China, this study integrates long-term Landsat time-series imagery (1990–2025) with deep learning to quantify both mangrove extent change and canopy degradation. To mitigate tidal inundation effects, a NDVI Pseudo-P75 compositing strategy was applied using Google Earth Engine (GEE), enabling consistent observation of mangrove canopies across tidal stages. Global Mangrove Watch v4 (GMW-V4) and HGMF2020 mangrove dataset for China were used as reference labels to train a ResNet34–UNet segmentation framework incorporating Digital Elevation Model (DEM) constraints. The model achieved high classification performance, with an IoU of 0.822 for mangroves and 0.981 for background, yielding a mean IoU of 0.902. The resulting maps, following manual verification, provided a robust basis for spatiotemporal and degradation analyses. Canopy condition was further assessed using the Enhanced Vegetation Index (EVI), which is less prone to saturation in high-biomass mangrove stands. Results show that mangrove area in the Nanliu River Delta expanded from 266 ha in 1990 to 1414 ha in 2025, with the annual expansion rate after 2005 being nearly seven times higher than that before 2005. Despite this net gain, a cumulative loss of 347.45 ha was recorded, primarily during 1990–2000, with approximately 70% converted to aquaculture and coastal engineering. Spatial analysis revealed that mangrove expansion occurred predominantly seaward, whereas both mangrove loss and canopy degradation exhibited an inverse J-shaped relationship with seawall proximity. More than 80% of mangrove loss occurred within 200 m of seawalls, indicating concentrated anthropogenic encroachment, while 75.6% of canopy degradation was observed within 350 m, potentially reflecting landward forest senescence. These results indicate a transition in dominant threats from permanent land conversion in the late 20th century to more subtle, internal functional degradation in recent decades, underscoring the need to complement extent-based assessments with canopy condition monitoring in mangrove conservation and management. Full article

Reliable traffic sign detection is essential for the safety of autonomous driving systems. However, manually annotating large-scale datasets for this task is resource-intensive, making semi-supervised learning (SSL) a vital alternative. Despite their potential, current SSL methods often struggle with unreliable pseudo-label filtering and limited detection accuracy. To address these limitations, we propose a novel framework integrating a Dual Confidence Fusion (DC-Fusion) module and a Structured Block-Regularized Neck (SBR-Neck). The former improves pseudo-label reliability by combining classification and localization confidence scores, while the latter optimizes feature representation through multi-scale fusion and block-wise regularization. To preserve high-frequency spatial details, SBR-Neck incorporates Spatial-Context-Aware Upsampling (SCA-Upsampling), which utilizes multi-granularity feature decomposition. Experimental results on a proprietary traffic sign dataset demonstrate that our method achieves ${\mathrm{mAP}}_{50}$ scores of 10.4%, 17.8%, 23.7%, and 32.1% using 1%, 2%, 5%, and 10% labeled data, respectively. These results surpass the “Efficient Teacher” baseline by margins ranging from 3.07% to 11%, confirming the framework’s ability to provide robust detection in complex traffic scenarios. Full article