IoT, Год журнала: 2025, Номер 6(1), С. 13 - 13
Опубликована: Фев. 10, 2025
Advances in deep learning (DL) models and next-generation edge devices enable real-time image classification, driving a transition from the traditional, purely cloud-centric IoT approach to edge-based AIoT, with cloud resources reserved for long-term data storage in-depth analysis. This innovation is transformative agriculture, enabling autonomous monitoring, localized decision making, early emergency detection, precise chemical application, thereby reducing costs minimizing environmental health impacts. The workflow of an AIoT system agricultural monitoring involves two main steps: optimal training tuning DL through extensive experiments on high-performance AI-specialized computers, followed by effective customization deployment advanced devices. review highlights key challenges practical applications, including: (i) limited availability data, particularly due seasonality, addressed public datasets synthetic generation; (ii) selection state-of-the-art computer vision algorithms that balance high accuracy compatibility resource-constrained devices; (iii) algorithm optimization integration hardware accelerators inference; (iv) recent advancements AI classification that, while not yet fully deployable, offer promising near-term improvements performance functionality.
Язык: Английский
Процитировано
2Plant Disease Phenotype Captioning Via Zero-Shot Learning with Semantic Correction Based on Llm
Опубликована: Янв. 1, 2025
Agriculture is the foundation of global food security and quality life, with staple crops such as rice, wheat, maize meeting dietary needs majority world's population. These are vulnerable to diseases that can cause severe yield losses; for instance, wheat rust disease results in annual losses exceeding \$2.9 billion. Accurate caption phenotypic characteristics plant play a crucial supportive role diagnosis, which essential ensure security. Existing methods agriculture fail adequately address heterogeneity visual phenotypes descriptions, resulting insufficient focus on characteristics. To mitigate this issue, we propose zero-shot image captioning framework, namely PDPC. PDPC uses an extensive descriptive corpus, syntactic analysis, optimization semantic structures significantly improve generalization descriptions. Additionally, have constructed dataset comprising 20,943 captions over 60 species 300 diseases. Experimental demonstrate framework outperforms existing models accurately describing disease. The introduction innovative not only improves accuracy descriptions but also provides robust support intelligent diagnosis management diseases, paving way improved health increased agricultural yields. Please click following link learn more about my paper: http://pdpc.samlab.cn/
Язык: Английский
Процитировано
0Integration of Deep Learning with Fox Optimization Algorithm for Early Detection and Classification of Tomato Leaf and Fruit Diseases
Engineering Technology & Applied Science Research, Год журнала: 2025, Номер 15(1), С. 19343 - 19348
Опубликована: Фев. 2, 2025
Tomato is a common vegetable crop extensively cultivated in the farming lands India. The hot climate of India perfect for its development, but particular weather conditions along with many other aspects affect growing tomato plants. Apart from these natural disasters and conditions, plant diseases consist major issue production. Precisely classifying leaf fruit plants vital step toward computerizing processes. Traditional disease detection models crops often fall short predictability. To address this, Machine Learning (ML) Deep (DL) have been developed, presenting advanced classification capabilities ability to manage vast variability agricultural data that conventional computer vision struggle with. This work presents an Integration DL Fox Optimization Algorithm (FOA) Recognition Classification Leaf Fruit Diseases (IDLFOA-DCTLFD). objective proposed IDLFOA-DCTLFD model enhance outcomes diseases. At initial stage, Median Filter (MF) used pre-processing Efficient Channel Attention-SqueezeNet (ECA-SqueezeNet) employed feature extraction. For hyperparameter tuning process, technique implements FOA. Finally, Wasserstein Generative Adversarial Network (WGAN) utilized method experimentally examined dataset. experimental validation methodology portrayed superior accuracy value 98.02%, surpassing existing techniques.
Язык: Английский
Процитировано
0CBSNet: An Effective Method for Potato Leaf Disease Classification
Plants, Год журнала: 2025, Номер 14(5), С. 632 - 632
Опубликована: Фев. 20, 2025
As potato is an important crop, disease detection and classification are of key significance in guaranteeing food security enhancing agricultural production efficiency. Aiming at the problems tiny spots, blurred edges, susceptibility to noise interference during image acquisition transmission leaf diseases, we propose a CBSNet-based recognition method. Firstly, convolution module called Channel Reconstruction Multi-Scale Convolution (CRMC) designed extract upper lower features by separating channel applying more optimized features, followed multi-scale operation capture changes effectively. Secondly, new attention mechanism, Spatial Triple Attention (STA), developed, which first reconstructs spatial dimensions input feature maps, then inputs reconstructed three types into each branches carries out targeted processing according importance thereby improving model performance. In addition, Bat–Lion Algorithm (BLA) introduced, combines Lion algorithm bat optimization makes process adaptive using adjust gradient direction updating algorithm. The BLA not only boosts model’s ability recognize but also ensures training stability enhances robustness handling noisy images. Experimental results showed that CBSNet achieved average Accuracy 92.04% Precision 91.58% on self-built dataset. It effectively extracts subtle spots blurry edges providing strong technical support for prevention control large-scale farming.
Язык: Английский
Процитировано
0Image-Based Blight Disease Detection in Crops using Ensemble Deep Neural Networks for Agricultural Applications
Journal of Natural Pesticide Research, Год журнала: 2025, Номер unknown, С. 100130 - 100130
Опубликована: Апрель 1, 2025
Язык: Английский
Процитировано
0Comparative Analysis of Cassava Leaf Disease Prediction Using the Deep Learning Approach
Опубликована: Июль 10, 2024
Язык: Английский
Процитировано
2Potato Leaf Disease Detection Based on a Lightweight Deep Learning Model
Machine Learning and Knowledge Extraction, Год журнала: 2024, Номер 6(4), С. 2321 - 2335
Опубликована: Окт. 14, 2024
Traditional methods of agricultural disease detection rely primarily on manual observation, which is not only time-consuming and labor-intensive, but also prone to human error. The advent deep learning has revolutionized plant by providing more accurate efficient solutions. management potato diseases critical the industry, as these can lead substantial losses in crop production. prompt identification classification leaf are essential mitigating such losses. In this paper, we present a novel approach that integrates lightweight convolutional neural network architecture, RegNetY-400MF, with transfer techniques accurately identify seven different types diseases. proposed method enhances precision reduces computational storage demands, mere 0.40 GFLOPs model size 16.8 MB. This makes it well-suited for use edge devices limited resources, enabling real-time environments. experimental results demonstrated accuracy identifying was 90.68%, comprehensive solution management.
Язык: Английский
Процитировано
1Plant Disease Recognition: A Comprehensive Mini Review
Опубликована: Июль 23, 2024
Язык: Английский
Процитировано
0