Optimizing Pear Leaf Disease Detection Through PL-DenseNet

Deleted Journal, Год журнала: 2025, Номер 67(1)

Опубликована: Фев. 1, 2025

Язык: Английский

---

Optimizing Deep Learning Models for Climate-Related Natural Disaster Detection from UAV Images and Remote Sensing Data

Journal of Imaging, Год журнала: 2025, Номер 11(2), С. 32 - 32

Опубликована: Янв. 24, 2025

This research study utilized artificial intelligence (AI) to detect natural disasters from aerial images. Flooding and desertification were two taken into consideration. The Climate Change Dataset was created by compiling various open-access data sources. dataset contains 6334 images UAV (unmanned vehicles) satellite then used train Deep Learning (DL) models identify disasters. Four different Machine (ML) used: convolutional neural network (CNN), DenseNet201, VGG16, ResNet50. These ML trained on our so that their performance could be compared. DenseNet201 chosen for optimization. All four performed well. ResNet50 achieved the highest testing accuracies of 99.37% 99.21%, respectively. project demonstrates potential AI address environmental challenges, such as climate change-related study’s approach is novel creating a new dataset, optimizing an model, cross-validating, presenting one DL detection. Three categories (Flooded, Desert, Neither). Our relates Environmental Sustainability. Drone emergency response would practical application project.

Язык: Английский

---

MoSViT: a lightweight vision transformer framework for efficient disease detection via precision attention mechanism

Frontiers in Artificial Intelligence, Год журнала: 2025, Номер 8

Опубликована: Март 26, 2025

Maize, a globally essential staple crop, suffers significant yield losses due to diseases. Traditional diagnostic methods are often inefficient and subjective, posing challenges for timely accurate pest management. This study introduces MoSViT, an innovative classification model leveraging advanced machine learning computer vision technologies. Built on the MobileViT V2 framework, MoSViT integrates CLA focus mechanism, DRB module, Block, LeakyRelu6 activation function enhance feature extraction accuracy while reducing computational complexity. Trained dataset of 3,850 images encompassing Blight, Common Rust, Gray Leaf Spot, Healthy conditions, achieves exceptional performance, with accuracy, Precision, Recall, F1 Score 98.75%, 98.73%, 98.72%, respectively. These results surpass leading models such as Swin Transformer V2, DenseNet121, EfficientNet in both parameter efficiency. Additionally, model's interpretability is enhanced through heatmap analysis, providing insights into its decision-making process. Testing small sample datasets further demonstrates MoSViT's generalization capability potential small-sample detection scenarios.

Язык: Английский

---

Optimized DINO model for accurate object detection of sesame seedlings and weeds

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Апрель 7, 2025

Abstract The application of intelligent agricultural machinery is crucial in modern production. However, environments where the target and surrounding morphology are highly similar, such as distinguishing sesame seedlings from weeds, problem essentially becomes one optimizing edge detection algorithms for similar targets. To address this issue object detection, we developed a custom dataset containing 1,300 images weeds. overcome high complexity low accuracy limitations original DINO model problem, backbone network was replaced with MobileNet V3, SENet attention mechanism neck structure were optimized, H-Swish6 activation function introduced to suit devices. Given higher degree lignification stems seedlings, these modifications improved overall Average Precision (AP) on COCO by 5.1% compared model. Specifically, $$\text {AP}_{S}$$ {AP}_{M}$$ increased 3.3% 3.8%, respectively, while {AP}_{50}$$ {AP}_{75}$$ 2.3% 3.2%. model’s parameter count reduced 29M, inference time lowered 60%, computational cost FLOPs decreased 43.72%. verify effectiveness improvements, On model, achieved maximum AP 81.8%, outperforming YOLOv7 5.6%, an FPS 24 frames per second. Ablation experiments verified improvements.However, aforementioned studies have not addressed scenarios targets domain.

Язык: Английский

---

Enhancing Rice Disease and Insect-Pest Detection through Augmented Deep Learning with Transfer Learning Techniques

Smart Agricultural Technology, Год журнала: 2025, Номер unknown, С. 100954 - 100954

Опубликована: Апрель 1, 2025

Язык: Английский

---

Deep Learning Techniques for Image Clustering and Classification

Advances in computational intelligence and robotics book series, Год журнала: 2025, Номер unknown, С. 37 - 62

Опубликована: Март 7, 2025

This chapter explores deep learning techniques for image clustering and classification, crucial tasks in computer vision. It discusses unsupervised methods supervised classification approaches, including traditional like k-means hierarchical clustering. The also highlights the transformative impact of CNNs, ensemble methods, transfer classification. uses case studies fruit agriculture brain tumor medical imaging to illustrate real-world applicability these models. Challenges data limitations, computational requirements, explainability are discussed. Future trends include self-supervised multimodal aims guide researchers practitioners leveraging effective across various domains.

Язык: Английский

---

Research on agricultural disease recognition methods based on very large Kernel convolutional network-RepLKNet

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 15, 2025

Agricultural diseases pose significant challenges to plant production. With the rapid advancement of deep learning, accuracy and efficiency disease identification have substantially improved. However, conventional convolutional neural networks that rely on multi-layer small-kernel structures are limited in capturing long-range dependencies global contextual information due their constrained receptive fields. To overcome these limitations, this study proposes a recognition method based RepLKNet, architecture with large kernel designs significantly expand field enhance feature representation. Transfer learning is incorporated further improve training model performance. Experiments conducted Plant Diseases Training Dataset, comprising 95,865 images across 61 categories, demonstrate effectiveness proposed method. Under five-fold cross-validation, achieved an overall (OA) 96.03%, average (AA) 94.78%, Kappa coefficient 95.86%. Compared ResNet50 (OA: 95.62%) GoogleNet 94.98%), demonstrates competitive or superior Ablation experiments reveal replacing kernels 3×3 5×5 convolutions results reductions up 1.1% OA 1.3% AA, confirming design. These robustness capability RepLKNet tasks.

Язык: Английский

---

RP-DETR: end-to-end rice pests detection using a transformer

Plant Methods, Год журнала: 2025, Номер 21(1)

Опубликована: Май 17, 2025

Язык: Английский

---

RLDD-YOLOv11n: Research on Rice Leaf Disease Detection Based on YOLOv11

Agronomy, Год журнала: 2025, Номер 15(6), С. 1266 - 1266

Опубликована: Май 22, 2025

Rice disease identification plays a critical role in ensuring yield stability, enabling precise prevention and control, promoting agricultural intelligence. However, existing approaches rely heavily on manual inspection, which is labor-intensive inefficient. Moreover, the significant variability features poses further challenges to accurate recognition. To address these issues, this paper proposes novel rice leaf detection model—RLDD-YOLOv11n. First, improved RLDD-YOLOv11n integrates SCSABlock residual attention module into neck layer enhance multi-semantic information fusion, thereby improving capability for small targets. Second, recognizing limitations of native upsampling YOLOv11n reconstructing rice-disease-related features, CARAFE incorporated. Finally, dataset focusing three common diseases—Bacterial Blight, Blast, Brown Spot—was constructed. The experimental results demonstrate effectiveness proposed improvements. achieved mean Average Precision (mAP) 88.3%, representing 2.8% improvement over baseline model. Furthermore, compared with mainstream lightweight YOLO models, exhibits superior performance robustness.

Язык: Английский

---

ST-YOLO: a deep learning based intelligent identification model for salt tolerance of wild rice seedlings

Frontiers in Plant Science, Год журнала: 2025, Номер 16

Опубликована: Июнь 2, 2025

Background In response to the limited models for salt tolerance detection in wild rice, subtle leaf features, and difficulty capturing stress characteristics, resulting low recognition rates accuracy, a deep learning-based ST-YOLO rice seedling phenotype evaluation identification model is proposed. Method order improve accuracy achieve lightweighting, multi branch structure DBB (Diverse Branch Block) used replace convolutional layers C2f module, reparameterization module proposed some modules. Diversified feature extraction paths are introduced enhance ability of extraction; Introducing CAFM (Context Aware Feature Modulation) convolution attention fusion modules into backbone network representation capabilities while improving features at various scales; Design more flexible effective spatial pyramid pooling layer using deformable information enhancement model’s represent target accuracy. Results The experimental results show that improved algorithm improves average precision by 2.7% compared with original network; rate 3.5%; recall 4.9%. Conclusion significantly current mainstream model, evaluates level varieties, screens out total 2 varieties extremely tolerant 7 tolerant, which meets real-time requirements, has certain reference value practical application.

Язык: Английский

---