ADMNet: adaptive deformable convolution large model combining multi-level progressive fusion for Building Change Detection

Geo-spatial Information Science, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 18

Published: Jan. 17, 2025

Building Change Detection (BCD) based on high-resolution Remote Sensing Images (RSI) simplifies urban surface monitoring. Nevertheless, the mainstream detection methods utilizing traditional convolution and attention mechanisms are often prone to errors due loss of edge detail information underutilization global context information. To address these issues, this paper presents a large model, namely ADMNet, which is built adaptive deformable designed handles various types building change First, we propose Siamese neural network (ADC) modules. The ADC module incorporates spatial offset parameters into convolutional kernel sampling mapping weights capture irregularly varying features for local receptive fields. Second, utilize model semantically driven enhance awareness construct long-range feature dependencies from multi-scale information, then integrated with locally structure achieve accurate localization. Furthermore, design Multi-Level Progressive Feature Fusion (MLPFF) that enhances characterization capabilities ensure internal integrity improves performance by integrating priori knowledge large-model transfer learning. evaluate effectiveness generalizability conduct comparative experiments current two datasets, LEVIR-CD WHU-CD, land cover dataset, SYSU-CD. results show ADMNet outperforms all methods. source code available at https://github.com/spaceYu180/ADMNet.

Language: Английский

---

A Domain-Adaptive Segmentation Method Based on Segment Anything Model for Mechanical Assembly

Measurement, Journal Year: 2024, Volume and Issue: 235, P. 114901 - 114901

Published: May 12, 2024

Language: Английский

---

Semantic segmentation method of photovoltaic cell microcracks based on EL polarization imaging

Solar Energy, Journal Year: 2025, Volume and Issue: 291, P. 113364 - 113364

Published: March 4, 2025

Language: Английский

---

An Improved V-Net Model for Thyroid Nodule Segmentation

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(7), P. 3873 - 3873

Published: April 1, 2025

Early diagnosis of increasingly common thyroid nodules is crucial for effectively and accurately managing the disease’s monitoring treatment process. In practice, manual segmentation methods based on ultrasound images are widely used; however, owing to limitations arising from imaging sources differences radiologist opinions, their standalone use may not be sufficient nodule segmentation. Therefore, there a growing focus developing automatic diagnostic approaches assist radiologists in diagnosis. Although current have yielded successful results, more research needed detection because complexity region, irregular tissues, blurred boundaries. This study proposes an improved V-Net model fully convolutional neural networks (V-Net) squeeze-and-excitation (SE) mechanisms detecting two-dimensional image data. addition strengths approach proposed model, mechanism was used emphasize important features suppress irrelevant by assigning weights significant model. Experimental studies utilized Digital Database Thyroid Image (DDTI) Nodule 3493 (TN3K) datasets, V-Net-based validated using V-Net, fusion SEV-Net methods. The results obtained experimental demonstrate that outperforms models, with Dice score 84.51% IoU 76.27% DDTI dataset. Similarly, TN3K dataset, it achieved superior performance compared all benchmarked scores 83.88% 75.50%, respectively. When considering context literature, demonstrated best among achieving average 80.39% dataset 79.69% according both metrics. 84.51%, competes at competitive level Ska-Net, which exhibits this metric 84.98% whereas existing models 75.5% achievement make effective tool can detection.

Language: Английский

---

Plant Disease Detection with Transfer Learning: Evaluating the Impact of Various Loss Functions and Explainable AI

Journal of Agricultural Biological and Environmental Statistics, Journal Year: 2025, Volume and Issue: unknown

Published: April 21, 2025

Language: Английский

---

Multi-objective evolutionary neural architecture search for medical image analysis using transformer and large language models in advancing public health

Applied Soft Computing, Journal Year: 2025, Volume and Issue: unknown, P. 113279 - 113279

Published: May 1, 2025

Language: Английский

---

Enhanced Pelvic CT Segmentation via Deep Learning: A Study on Loss Function Effects

Deleted Journal, Journal Year: 2025, Volume and Issue: unknown

Published: May 28, 2025

Effective radiotherapy planning requires precise delineation of organs at risk (OARs), but the traditional manual method is laborious and subject to variability. This study explores using convolutional neural networks (CNNs) for automating OAR segmentation in pelvic CT images, focusing on bladder, prostate, rectum, femoral heads (FHs) as an efficient alternative segmentation. Utilizing Medical Open Network AI (MONAI) framework, we implemented compared U-Net, ResU-Net, SegResNet, Attention U-Net models explored different loss functions enhance accuracy. Our involved 240 patients prostate 220 other organs. The models' performance was evaluated metrics such Dice similarity coefficient (DSC), Jaccard index (JI), 95th percentile Hausdorff distance (95thHD), benchmarking results against expert masks. SegResNet outperformed all models, achieving DSC values 0.951 0.829 0.860 0.979 left FH, 0.985 right FH (p < 0.05 vs. ResU-Net). also excelled, particularly bladder rectum Experiments with showed that consistently delivered optimal or equivalent across OARs, while DiceCE slightly enhanced (DSC = 0.845, p 0.0138). These indicate advanced CNNs, especially paired optimized functions, provide a reliable, methods, promising improved precision planning.

Language: Английский

---

Liver image segmentation method based on improved EfficientNet-B0

Journal of Computational Methods in Sciences and Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: May 27, 2025

This paper presents a liver image segmentation method based on the EABRDeNet model, which incorporates architectural concept of U-Net. In encoding stage, EfficientNet-B0 serves as backbone network, combined with SEBlock mechanism and residual blocks to enhance extraction key features address gradient-related issues. decoding deconvolution upsampling are used for precise segmentation. Liver images from open-source websites pre-processed data-augmented construct G dataset. The model is trained this dataset, results obtained through weight-sharing mechanism. To verify model’s effectiveness, comparative experiments conducted dataset U-Net + Dice_Focal_Loss. Metrics such accuracy, loss, Dice coefficient evaluation. experimental show that outperforms other two models. Specifically, training set, has an average loss 0.0025979, accuracy 0.9876249, 0.985658. On test 0.0029967, 0.987649, 0.9845206. contrast, Dice_Focal_Loss have relatively higher losses lower accuracies coefficients, indicating better performance stability in tasks.

Language: Английский

---

CTDUNet: A Multimodal CNN–Transformer Dual U-Shaped Network with Coordinate Space Attention for Camellia oleifera Pests and Diseases Segmentation in Complex Environments

Plants, Journal Year: 2024, Volume and Issue: 13(16), P. 2274 - 2274

Published: Aug. 15, 2024

is a crop of high economic value, yet it particularly susceptible to various diseases and pests that significantly reduce its yield quality. Consequently, the precise segmentation classification diseased Camellia leaves are vital for managing effectively. Deep learning exhibits significant advantages in plant pests, complex image processing automated feature extraction. However, when employing single-modal models segment

Language: Английский

---

Brain tumor segmentation algorithm based on pathology topological merging

Multimedia Tools and Applications, Journal Year: 2024, Volume and Issue: 83(40), P. 88019 - 88037

Published: March 21, 2024

Language: Английский

---