Apnet: Lightweight network for apricot tree disease and pest detection in real-world complex backgrounds

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

Опубликована: Янв. 9, 2025

Язык: Английский

---

Enhanced hybrid attention deep learning for avocado ripeness classification on resource constrained devices

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

Опубликована: Янв. 29, 2025

Attention mechanisms such as the Convolutional Block Module (CBAM) can help emphasize and refine most relevant feature maps color, texture, spots, wrinkle variations for avocado ripeness classification. However, CBAM lacks global context awareness, which may prevent it from capturing long-range dependencies or patterns relationships between distant regions in image. Further, more complex neural networks improve model performance but at cost of increasing number layers train parameters, not be suitable resource constrained devices. This paper presents Hybrid Neural Network (HACNN) classifying on It aims to perform local enhancement capture relationships, leading a comprehensive extraction by combining attention modules models. The proposed HACNN combines transfer learning with hybrid mechanisms, including Spatial, Channel, Self-Attention Modules, effectively intricate features fourteen thousand images. Extensive experiments demonstrate that EfficienctNet-B3 significantly outperforms conventional models regarding accuracy 96.18%, 92.64%, 91.25% train, validation, test models, respectively. In addition, this consumed 59.81 MB memory an average inference time 280.67 ms TensorFlow Lite smartphone. Although ShuffleNetV1 (1.0x) consumes least resources, its testing is only 82.89%, insufficient practical applications. Thus, MobileNetV3 Large exciting option has 91.04%, usage 26.52 MB, 86.94 These findings indicated method enhances classification ensures feasibility implementation low-resource environments.

Язык: Английский

---

A hybrid deep learning model approach for automated detection and classification of cassava leaf diseases

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

Опубликована: Фев. 27, 2025

Detecting cassava leaf disease is challenging because it hard to identify diseases accurately through visual inspection. Even trained agricultural experts may struggle diagnose the correctly which leads potential misjudgements. Traditional methods these are time-consuming, prone error, and require expert knowledge, making automated solutions highly preferred. This paper explores application of advanced deep learning techniques detect as well classify includes EfficientNet models, DenseNet169, Xception, MobileNetV2, ResNet Vgg19, InceptionV3, InceptionResNetV2. A dataset consisting around 36,000 labelled images leaves, afflicted by such Cassava Brown Streak Disease, Mosaic Green Mottle, Bacterial Blight, healthy was used train models. Further were pre-processed converting them into grayscale, reducing noise using Gaussian filter, obtaining region interest Otsu binarization, Distance transformation, Watershed technique followed employing contour-based feature selection enhance model performance. Models, after fine-tuned with ADAM optimizer computed that among tested hybrid (DenseNet169 + EfficientNetB0) had superior performance classification accuracy 89.94% while EfficientNetB0 highest values precision, recall, F1score 0.78 each. The novelty lies in its ability combine DenseNet169's reuse capability EfficientNetB0's computational efficiency, resulting improved scalability. These results highlight for accurate scalable diagnosis, laying foundation plant monitoring systems.

Язык: Английский

---

Enhancement of tea leaf diseases identification using modified SOTA models

Neural Computing and Applications, Год журнала: 2024, Номер unknown

Опубликована: Дек. 5, 2024

Язык: Английский

---

RTR_Lite_MobileNetV2: A Lightweight and Efficient Model for Plant Disease Detection and Classification

Current Plant Biology, Год журнала: 2025, Номер unknown, С. 100459 - 100459

Опубликована: Фев. 1, 2025

Язык: Английский

---

Automated grading of oleaster fruit using deep learning

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

Опубликована: Фев. 12, 2025

The agriculture sector is crucial to many economies, particularly in developing regions, with post-harvest technology emerging as a key growth area. oleaster, valued for its nutritional and medicinal properties, has traditionally been graded manually based on color appearance. As global demand rises, there growing need efficient automated grading methods. Therefore, this study aimed develop real-time machine vision system classifying oleaster fruit at various velocities. Initially, the offline phase, dataset containing video frames of four different quality classes categorized Iranian national standard, was acquired linear conveyor belt velocities (ranging from 4.82 21.51 cm/s). Mask R-CNN algorithm used segment extracted obtain position boundary samples. Experimental results indicated that, 100% detection rate an average instance segmentation accuracy error ranging 4.17 5.79%, capable accurately segmenting all examined velocity levels. fivefold cross validation that general YOLOv8x YOLOv8n models, created using obtained levels, have similarly reliable classification performance. given simpler architecture lower processing time requirements, model evaluate mode. overall 92%, sensitivity range 87.10–94.89% distinguishing cm/s. demonstrate effectiveness deep learning-based models machines fruit.

Язык: Английский

---

BiFPN-enhanced SwinDAT-based cherry variety classification with YOLOv8

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

Опубликована: Фев. 13, 2025

Accurate classification of cherry varieties is crucial for their economic value and market differentiation, yet genetic diversity visual similarity make manual identification challenging, hindering efficient agricultural trade practices. This study addresses this issue by proposing a novel deep learning-based hybrid model that integrates BiFPN with the YOLOv8n-cls framework, enhanced Swin Transformer Deformable Attention (DAT) techniques. The was trained evaluated on newly constructed dataset comprising from Turkey's Western Mediterranean region. Experimental results demonstrated effectiveness proposed approach, achieving precision 91.91%, recall 92.0%, F1-score 91.93%, an overall accuracy 91.714%. findings highlight model's potential to optimize harvest timing, ensure quality control, support export classification, thereby contributing improved practices outcomes.

Язык: Английский

---

Mango fruit diseases severity estimation based on image segmentation and deep learning

Deleted Journal, Год журнала: 2025, Номер 7(2)

Опубликована: Фев. 15, 2025

Plant disease severity is the ratio between surface area of symptoms and total plant unit (e.g. fruit, leaf). It related to diagnosis has several advantages for farmers. therefore a key element in protection management diseases. In literature, there are three proposed categories determination solutions: solutions based on segmentation algorithms, classical ML algorithms those DL algorithms. Despite their advantages, these have number limitations, including i) subjectivity data labeling, ii) loss information lesion contours during (manual) iii) focused estimating from leaves, although diseases can also affect other parts plant, such as fruits. this paper, we present solution four mango fruit diseases, namely Alternariose, Anthracnose, Aspergillus rot Stem rot. This ResNet50 CNN uses dataset automatically labeled by algorithm two image color space thresholding. The achieved an accuracy, precision F1_score 97.82%, 97.09% 97.79%, respectively, test then deployed mobile application with diagnostic previously proposed. will help growers, particularly Sahelian countries like Senegal, manage earlier. article presents not but fruit. label used, order avoid any errors that might result manual labeling. deep learning, particular model, estimate (alternaria, anthracnose, aspergillus stem rot) according stages: healthy, early, intermediate final. finally growers.

Язык: Английский

---

Artificial intelligence for sustainable farming with dual branch convolutional graph attention networks in rice leaf disease detection

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

Опубликована: Март 27, 2025

Rice is susceptible to various diseases, including brown spot, hispa, leaf smut, bacterial blight, and blast, all of which can negatively impact crop yields. Current disease detection methods encounter several challenges, such as reliance on a single dataset that diminishes accuracy, the use complex models, limitations posed by small datasets hinder performance. To overcome these this paper presents novel hybrid deep learning (DL) approach for classifying rice diseases. The proposed model leverages two distinct datasets: Leaf Diseases Dataset Disease Images Dataset. It enhances image quality through advanced techniques: Upgraded Weighted Median Filtering (Up-WMF) minimize noise Aligned Gamma-based Contrast Limited Adaptive Histogram Equalization (AG-CLAHE) improve contrast. Features from images are extracted using Discrete Wavelet Transform (DWT), Gray Level Run Length Matrix (GLRLM), learning-based VGG19 features. optimize performance, most significant features selected Bio-Inspired Artificial Hummingbird (BI-AHB) method, streamlines complexity. Classification diseases conducted new known Dual Branch Convolutional Graph Attention Neural Network (DB-CGANNet). This demonstrates remarkable achieving 98.9% accuracy 99.08% image, surpassing existing techniques. methodology facilitating improved management crops contributing increased agricultural productivity.

Язык: Английский

---

A compact deep learning approach integrating depthwise convolutions and spatial attention for plant disease classification

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

Опубликована: Апрель 2, 2025

Abstract Plant leaf diseases significantly threaten agricultural productivity and global food security, emphasizing the importance of early accurate detection effective crop health management. Current deep learning models, often used for plant disease classification, have limitations in capturing intricate features such as texture, shape, color leaves. Furthermore, many these models are computationally expensive less suitable deployment resource-constrained environments farms rural areas. We propose a novel Lightweight Deep Learning model, Depthwise Separable Convolution with Spatial Attention (LWDSC-SA), designed to address enhance feature extraction while maintaining computational efficiency. By integrating spatial attention depthwise separable convolution, LWDSC-SA model improves ability detect classify diseases. In our comprehensive evaluation using PlantVillage dataset, which consists 38 classes 55,000 images from 14 species, achieved 98.7% accuracy. It presents substantial improvement over MobileNet by 5.25%, MobileNetV2 4.50%, AlexNet 7.40%, VGGNet16 5.95%. validate its robustness generalizability, we employed K-fold cross-validation K=5, demonstrated consistently high performance, an average accuracy 98.58%, precision 98.30%, recall 98.90%, F1 score 98.58%. These results highlight superior performance proposed demonstrating outperform state-of-the-art terms remaining lightweight efficient. This research offers promising solution real-world applications, enabling resource-limited settings contributing more sustainable practices.

Язык: Английский

---