SCIFL-Net: Seam carving-based image forgery localization network

Hybrid multi-scale CNN-Transformer network for structural surface crack segmentation

GAST-NET: A multi-modal and multi-task deep learning framework for preoperative prediction of perineural invasion and prognostic risk in gastric cancer.

Unsupervised domain adaptation for medical image segmentation by pre-training and 3D multi-scale feature fusion

Diffusion-based image captioning models effectively mitigate the token dependency issue inherent in autoregressive methods. However, the noise introduced in diffusion methods weakens sentence information, resulting in insufficient ability of image-text feature alignment. To address this issue, we propose a multipath alignment learning for image captioning with feature mask (MALFM-Captioner) method. Leveraging both global and regional visual features, we first introduce a feature masked module (FMM) that enables the model to reconstruct masked visual information during training, thereby enhancing its capability to learn discriminative visual representations. Concurrently, we perform cross-attention between deep image and text features and fuse the outputs via weighted summation, effectively mitigating the image-text misalignment problem inherent to single-path paradigms. Furthermore, we design a concise yet effective gated feature fusion module (GFFM) to integrate complementary feature alignment results, improving the accuracy and semantic fidelity of generated captions. Extensive experiments on the MS COCO and Flickr 30K datasets demonstrate that MALFM-Captioner achieves 0.9% and 1.9% improvements in Bleu-4 and CIDEr metrics, respectively, and exhibits competitive performance against state-of-the-art models (e.g., DDCap and Bit Diffusion).

Industrial anomaly detection and localization have become key procedures in modern manufacturing for product quality assurance. However, it is still challenging for the emerging deep learning-based unsupervised anomaly detection approaches to simultaneously realize high-precision anomaly detection and localization, and generate high-quality anomaly masks end-to-end. Meanwhile, the limited number of normal samples and random nature of anomalies leads to insufficient discriminative information available for model training. To this end, a novel knowledge-prior memory discrimination based generative adversarial network (KMDGAN) is demonstrated for high-performance industrial anomaly detection and localization with merely few anomaly samples utilized. KMDGAN leverages both normal and anomaly features to learn feature-level spatial relationships and generate accurate anomaly masks end-to-end. Firstly, based on the idea of knowledge-prior discrimination, a memory feature matching module is elaborately designed to construct a memory knowledge base of normal product features, and considers the positive and negative difference information from the spatial location features, promoting the detection and localization of abnormal regions. Secondly, a cross-scale feature fusion module is developed to solve the multi-scale feature redundancy problem via attention optimization, while an anomaly map modulation technique enhances the integration of global context and local details. Finally, an edge-context hybrid module is designed to address the limitation of single information and enhance the perceptual comprehension ability by fully extracting the edge and overall information of anomaly sample features. Experimental results on the MVTec AD and VisA datasets demonstrate that our KMDGAN model achieves a salient anomaly detection and localization performance, outperforming the existing advanced counterparts.

A computational framework for Alzheimer's disease detection using SwinRes Transformer.

Hyperparathyroidism (HPT) and thyroid nodule (TN) are caused by abnormalities in the parathyroid and thyroid glands, respectively. Due to their proximity, small size, and similar ultrasound characteristics, the traditional object detection algorithms often struggle to accurately differentiate between HPT and TN when both lesions coexist in ultrasound images, leading to a high rate of misdiagnosis. In order to achieve accurate detection of HPT and TN when the two lesions coexist, we constructed three comprehensive object detection datasets: one containing only hyperparathyroidism (HPTD), one containing only thyroid nodules (TND), and one mixed dataset that includes both types of lesions (HPT-TND). A novel multi-branch feature fusion DETR network (MB-DETR) is proposed based on the Real-Time Detection Transformer (RT-DETR) model. We redesigned the feature fusion module and incorporated asymmetric convolution to enhance feature extraction. To validate the proposed MB-DETR performance, the experiments have been carried out on the three datasets. Our model achieved a superior performance compared to the state-of-the-art object detection models in the key metrics such as F1, Precision, and Recall, while significantly reducing computational costs. Additionally, the ablation studies confirmed the effectiveness of asymmetric convolution and the multi-branch feature fusion module in terms of enhancement of detection performance. The experimental results show that the Multi-Branch Feature Fusion incorporated with the asymmetric convolution improves the local feature extraction capability of the DETR model. It is concluded that the proposed MB-DETR model outperforms the existing ones in the detection of TN and HPT when both lesions coexist and thus effectively assists in the diagnosis of the correlated disease.

Semi-supervised learning for medical image segmentation is often hindered by a conservative approach to unlabeled data, where information-rich yet uncertain regions-such as complex structures and ambiguous boundaries-are typically discarded to ensure pseudo-label quality. This "avoidance" strategy limits the model's ability to learn from the most challenging areas, which are critical for clinical accuracy. To address these limitations, we propose Semi-URF, an Uncertainty-Aware Region Filtering and Fusion framework. Instead of treating uncertainty as an obstacle to be avoided, Semi-URF progressively exploits it as a supervisory signal to enhance learning from unlabeled data. Our approach establishes a synergistic closed loop with three key innovations. First, an Uncertainty Distribution Adaptive Thresholding (UDAT) mechanism adaptively separates reliable from unreliable regions by tracking the model's evolving uncertainty distribution, maximizing the use of informative pixels from unlabeled data. Second, our Bidirectional Uncertainty-Consistent Exchange (BUCE) method facilitates learning from unreliable regions by exchanging patches between labeled and unlabeled data, subject to a semantic consistency constraint. Finally, to enhance the model's foundational capabilities for this task, a Frequency-Enhanced Feature Fusion (FEFF) module uses fast wavelet transforms and cross-attention to sharpen the perception of high-frequency boundary details. Comprehensive experiments demonstrate that Semi-URF outperforms state-of-the-art methods, especially in situations where annotated data are severely limited. By transforming uncertainty from a problem into a solution, Semi-URF significantly reduces reliance on costly expert annotations. This approach facilitates the development of more efficient and cost-effective AI models for clinical applications. The code is available at https://github.com/senseiyang/Semi-URF.

MedSAM-guided geometry-aware 2D-3D feature fusion for medical image registration.

Accurate extraction of mountaintop areas from grid digital elevation models (DEMs) is essential for terrain analysis, geomorphological research, hydrological modeling, natural disaster monitoring, and emergency communication site selection. However, existing deep-learning-based methods often suffer from inadequate representation of local details and limited global contextual awareness, leading to blurred boundaries and reduced segmentation accuracy in complex mountainous terrains. To address these limitations, this study proposes a dual-encoder and global-context-enhanced TransUNet framework, named DEGC-TransUNet, for automated mountaintop delineation. The architecture integrates a convolutional encoder to capture fine-grained local terrain features and a MaxViT-based encoder to model multi-scale global context by encoding low-dimensional topographic attributes such as slope and curvature. A dedicated feature fusion module harmonizes complementary representations from both encoding paths, while a BiFormer-based strategy is introduced at the bottleneck to strengthen long-range dependencies and enhance convergence. The experimental results demonstrate that DEGC-TransUNet significantly outperforms baseline models such as TransUNet, DE-TransUNet, and GC-TransUNet, with relative improvements of 19.8% in Intersection over Union (IoU), 10.4% in overall accuracy (ACC), and 10.9% in F1-score. These findings provide a robust solution for mountaintop extraction, with significant potential in analyzing geomorphological evolution, simulating soil erosion, modeling species distribution in “sky island” ecosystems, and optimizing strategic placements for communication base stations and wind energy infrastructures.

Marine ecosystem integrity is paramount to global stability. With the advancement of industrialization, various types of waste are discharged into the ocean, accumulating through the food chain and ultimately threatening human health and the global climate environment. To achieve precise and efficient cleanup of marine debris, traceability is essential, with detection and classification serving as critical steps. To address the issues of missed detection and occlusion caused by the irregular shapes of marine debris due to water pressure or structural characteristics, as well as the coexistence of multi-scale objects resulting from aggregation and shooting angles, this study proposes the MD-YOLO model based on the YOLOv11L architecture. Firstly, a deformable attention mechanism is introduced in the neck network to achieve dynamic sampling and precise localization of targets with imbalanced aspect ratios. Secondly, a context-aware multi-scale feature fusion module is embedded in the backbone network to effectively mitigate the issue of missed detection of small targets when objects of different sizes coexist. Finally, a cooperative spatial-channel attention mechanism is designed in the detection head to enhance the feature representation capability in visible regions and infer occluded areas, thereby significantly suppressing occlusion interference. Experiments conducted on a self-constructed dataset containing 5095 images demonstrate that the proposed method achieves 86.7% in mAP@0.5, 67.6% in mAP@0.5:0.95, and an F1 score of 0.83, significantly outperforming comparative methods. This study provides key technical support for the effective traceability of marine debris.

Accurate rice mapping is critical for grain yield estimation and food security, yet traditional methods often struggle with asynchronous data quality and the inherent statistical gap between SAR and optical signals. To bridge this gap, we propose MMDFRNet, a novel multi-modal deep learning framework that synergistically integrates Sentinel-1 SAR and Sentinel-2 optical imagery. Unlike conventional static fusion approaches, MMDFRNet features a dual-stream modality-specific encoder architecture designed to decouple structural backscattering signals from spectral reflectance. Central to this framework is the multi-modal feature fusion (MMF) module, which employs an adaptive attention mechanism to dynamically align and recalibrate features based on their reliability, effectively mitigating noise from compromised modalities. Additionally, a multi-scale feature fusion (MSF) module is incorporated to coordinate hierarchical semantic information, enhancing boundary delineation in fragmented landscapes. Extensive experiments conducted across multiple study areas in China demonstrate the superiority of MMDFRNet. The model achieves a Precision of 0.9234, an IoU of 0.8612, and an F1-score of 0.9252. Notably, it consistently outperforms state-of-the-art benchmarks (e.g., UNetFormer, STMA, and CCRNet) by margins of up to 11.72% (Precision) and 7.39% (IoU) compared to classic baselines. Furthermore, rigorous ablation studies and degradation analyses confirm the model’s robustness, verifying its ability to transform the degradation paradox into a performance booster through pixel-wise adaptive alignment. Consequently, MMDFRNet offers a promising solution for precise rice area statistics and long-term monitoring in complex agricultural landscapes.

Advancements in infrared imaging technology have introduced a novel perspective for inspecting power transmission lines. Nevertheless, the inherent low contrast and indistinct edges of infrared images present significant challenges, rendering the direct application of traditional semantic segmentation algorithms unsatisfactory. To mitigate this problem, we propose a multi-neighborhood densely connected network architecture. This framework incorporates two pivotal modules: the Multi-Head Squeeze-and-Excitation (MHSE) module and the Multi-Neighborhood Feature Fusion (MNFF) module. The MHSE enhances local feature representations by capturing nuanced feature interactions, thereby alleviating the issue of imbalanced global feature weight distribution. The MNFF aggregates feature data from multiple adjacent nodes at each node’s input, which not only facilitates the integration of multi-scale target features but also leverages neighborhood information to precisely localize and amplify features within specific regions. Furthermore, we have built the first Infrared Dataset of Power Transmission Line Suspension Clamp (CLAMPTISS) to substantiate our approach. Empirical evidence demonstrates that our proposed network surpasses state-of-the-art networks across three key metrics: the mean Intersection over Union (mIoU) and localization accuracy (Pd) have increased by 8.3% and 13.3%, respectively, while the false alarm rate (Fa) has decreased by 38.2%.

MAFNet: A multi-modal adaptive fusion network-based approach for individual building extraction from oblique photogrammetry

SegMeshNet: Joint heart segmentation and mesh reconstruction with task-aware shared attention.

iAFP-fLRM: Accurate identification of antifungal peptides via hybrid deep learning architecture and multi-modal feature fusion.

HighlightsWhat are the main findings?A lightweight RFD-BiSeNetV2 network is developed for accurate floodwater segmentation.The proposed model achieves a high mean Intersection over Union (mIoU) of 97.10% while enabling real-time inference.What are the implications of the main findings?The model can be effectively deployed in real-time flood monitoring systems.The framework provides a practical solution for rapid flood detection and environmental disaster management.Flood disasters pose significant threats to human life and infrastructure, creating an urgent need for reliable vision-based environmental sensing technologies for rapid floodwater identification. Vision-based platforms such as unmanned surface vehicles (USVs) provide an effective solution for monitoring inland water environments; however, accurate floodwater segmentation remains challenging due to complex water boundaries, reflections, and background interference. To address these issues, we propose RFD-BiSeNet V2, a lightweight semantic segmentation network. Building upon BiSeNet V2, our model integrates an edge-aware learning strategy to track dynamic contours, a feature refinement module to suppress reflection noise, and a multi-scale feature fusion module to accommodate varying morphological scales. Evaluated on a comprehensive dataset incorporating USV data, UAV imagery, and diverse real-world scenes, RFD-BiSeNet V2 achieves an mIoU of 97.10%, outperforming the baseline by 6.68%. Crucially, the results demonstrate the practical implications of our architectural advancements: the edge-aware and feature refinement modules successfully sharpen ambiguous water boundaries and effectively filter out severe surface reflections, directly driving the segmentation accuracy. With a compact size of 5.95M parameters and real-time inference capabilities, the model offers a robust and highly efficient solution suitable for resource-constrained deployments across diverse intelligent environmental sensing systems.

Abstract Industrial weld seam image segmentation often suffers from accuracy limitations due to neglected linear geometric features and complex background interference. To address this, this paper proposes a weld-aware dual-attention U-Net model, named WADA-UNet. The architecture incorporates a weld-direction-aware spatial attention module and a weld-feature-enhanced channel attention module. These modules interact synergistically through a spatial-channel crossguidance mechanism embedded within the skip connections. Furthermore, an adaptive multi-scale feature fusion module is introduced to dynamically adjust fusion weights according to weld dimensions, thereby improving adaptability to variations in weld width. This paper also adopts a collaborative training strategy to further enhance model performance, which integrates multi-task loss functions, knowledge distillation, and probabilistic calibration. Experimental results show that WADA-UNet achieves a Dice coefficient of 0.9125, a Mean Intersection over Union (MIOU) of 0.9167, and a Mean Pixel Accuracy (MPA) of 0.9592, outperforming all benchmark methods. These results con-firm the model's strong generalization ability and robustness, highlighting its potential as a reliable solution for auto-mated welding quality inspection.

BackgroundIndustrial weld defect detection is challenged by the minimal grayscale contrast between defects and the background, as well as by blurred defect edges, which together hinder the performance of detection algorithms. Moreover, practical industrial environments require high detection accuracy, fast inference speed, and flexible deployment.ObjectiveTo address these challenges, this study proposes an improved YOLOv8n defect detection method that enables more accurate, faster, and lightweight automated weld defect detection.MethodsThe key improvements are as follows. First, in the backbone, the original C2f module is replaced by the C2f_OREPA feature extraction module, constructed with the Online Convolution Parameterization Approach (OREPA), which reduces computational complexity and enhances feature representation. Second, a downsampling module, DCDConv, is introduced to replace the conventional convolution after the first standard convolution layer, allowing better preservation of fine defect features and improving the detection of subtle defects. Additionally, in the neck, a cross-scale feature fusion module (CCFM) is incorporated to improve detection performance across defects of different scales.ResultsExperiments on our self-constructed dataset comprising eight weld defect categories show that the improved model achieves a mean average precision (mAP) of 87.6%, a 4.5% increase over the original YOLOv8n. Meanwhile, the model reduces the number of parameters by 26.9%, decreases computational cost by 35.7%, and achieves an inference speed of 103 frames per second (FPS). On the public NEU-DET dataset, the improved model obtains an mAP of 82.8%, outperforming the original YOLOv8n by 6.7%. Overall, the proposed model surpasses mainstream object detection frameworks, including YOLOv8n, YOLOv12n, Faster R-CNN, and RetinaNet.ConclusionIn summary, the proposed method provides an accurate, efficient, and deployment-friendly solution for weld defect detection in industrial applications, demonstrating substantial practical value.