A high-throughput ResNet CNN approach for automated grapevine leaf hair quantification

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: Jan. 10, 2025

Abstract The hairiness of the leaves is an essential morphological feature within genus Vitis that can serve as a physical barrier. A high leaf hair density present on abaxial surface grapevine influences their wettability by repelling forces, thus preventing pathogen attack such downy mildew and anthracnose. Moreover, hairs favorable habitat may considerably affect abundance biological control agents. unavailability accurate efficient objective tools for quantifying makes study intricate challenging. Therefore, validated high-throughput phenotyping tool was developed established in order to detect quantify using images single discs convolution neural networks (CNN). We trained modified ResNet CNNs with minimalistic number efficiently classify area covered hairs. This approach achieved overall model prediction accuracy 95.41%. As final validation, 10,120 input from segregating F1 biparental population were used evaluate algorithm performance. CNN-based phenotypic results compared ground truth data received two experts revealed strong correlation R values 0.98 0.92 root-mean-square error 8.20% 14.18%, indicating performance consistent expert evaluations outperforms traditional manual rating. Additional validation between vs. non-expert six varieties showed non-experts contributed over- underestimation trait, absolute 0% 30% -5% -60%, respectively. Furthermore, panel 16 novice evaluators produced significant bias set varieties. Our provide clear evidence need hairiness.

Language: Английский

---

Hyperparameter Tuning for Plant Leaf Disease Detection Using Convolutional Neural Networks

SSRN Electronic Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

Language: Английский

---

Evaluation of Different Few-Shot Learning Methods in the Plant Disease Classification Domain

Biology, Journal Year: 2025, Volume and Issue: 14(1), P. 99 - 99

Published: Jan. 19, 2025

Early detection of plant diseases is crucial for agro-holdings, farmers, and smallholders. Various neural network architectures training methods have been employed to identify optimal solutions disease classification. However, research applying one-shot or few-shot learning approaches, based on similarity determination, the plantdisease classification domain remains limited. This study evaluates different loss functions used in learning, including Contrastive, Triplet, Quadruplet, SphereFace, CosFace, ArcFace, alongside various backbone networks, such as MobileNet, EfficientNet, ConvNeXt, ResNeXt. Custom datasets real-life images, comprising over 4000 samples across 68 classes diseases, pests, their effects, were utilized. The experiments evaluate standard transfer approaches two function. Results demonstrate superiority cosine-based Siamese networks embedding extraction Effective model organization are determined. Additionally, impact data normalization tested, generalization ability models assessed using a special dataset consisting 400 images difficult-to-identify cases.

Language: Английский

---

From Pixels to Protection: Deep Learning Approaches for Plant Leaf Disease Detection

Deleted Journal, Journal Year: 2025, Volume and Issue: 3(1), P. 8 - 14

Published: Jan. 24, 2025

Plants are essential for human survival. However, diseases affecting plant leaves can lead to significant reductions in crop yield and economic losses. Detecting these early is crucial agriculture. To overcome limitations, machine learning has been employed automate the identification of leaf diseases. By analysing features such as colour, intensity, shape, models classify into specific categories, offering faster more accurate results than conventional approaches. Various ML techniques used identify leaves, with deep gaining attention its ability perform advanced feature extraction. CNNs have become a highly effective tool disease identification, thanks their automatically extract from images achieve high classification accuracy. Their hierarchical structure enables them detect simple patterns initial layers progressively learn complex deeper layers, capturing intricate details symptoms. Additionally, process large datasets multiple accurately, even limited labelled data, by leveraging pre-trained through transfer learning.

Language: Английский

---

Enhanced Tomato Leaf Disease Classification and Localization using Advanced Feature Extraction and Transfer Learning

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 5, 2025

Tomato plants are susceptible to various diseases that significantly impact crop yield and quality. Accurate timely identification of these is crucial for effective management mitigation. This study presents a deep learning-based methodology enhancing disease prediction, classification, precise localization affected areas within tomato leaves. The proposed approach leverages combination statistical, texture (Tamura GLCM), geometry, color features extracted from leaf images. To further enrich feature representation, wavelet analysis employed. model not only classifies ten prevalent but also estimates the proportion area, providing valuable insights severity assessment. Evaluated on dataset comprising 10,000 images, our achieves remarkable accuracy 99.50%. robust performance underscores efficacy in accurate diagnosis, benefitting farmers researchers by enabling prompt intervention efficient strategies.

Language: Английский

---

Revolutionizing Agriculture With Automated Plant Disease Detection

IGI Global eBooks, Journal Year: 2025, Volume and Issue: unknown, P. 267 - 296

Published: Feb. 7, 2025

Automated plant disease detection using computer vision has transformed agriculture by addressing challenges in health management, productivity, and sustainability. This chapter explores advancements from traditional methods to AI-enhanced deep learning multi-modal imaging, enabling early detection, real-time processing, precise interventions. Applications like precision agriculture, IoT integration, data-driven decision-making foster eco-friendly practices resource efficiency. Despite such as data quality, scalability, accessibility, future innovations collection, sustainable hardware, collaboration promise shape resilient agricultural systems. By aligning technology with sustainability, automated supports food security, environmental conservation, the evolution of modern farming practices.

Language: Английский

---

A Comparison of Different Deep Learning Models for Plant Leaf Disease Detection

Communications in computer and information science, Journal Year: 2025, Volume and Issue: unknown, P. 81 - 92

Published: Jan. 1, 2025

Language: Английский

---

A Lightweight Tea Bud-Grading Detection Model for Embedded Applications

Agronomy, Journal Year: 2025, Volume and Issue: 15(3), P. 582 - 582

Published: Feb. 26, 2025

The conventional hand-picking of tea buds is inefficient and leads to inconsistent quality. Innovations in bud identification automated grading are essential for enhancing industry competitiveness. Key breakthroughs include detection accuracy lightweight model deployment. Traditional image recognition struggles with variable weather conditions, while high-precision models often too bulky mobile applications. This study proposed a YOLOV5 model, which was tested on three types across different scenarios. It incorporated convolutional network compact feature extraction layer, significantly reduced parameter computation. achieved 92.43% precision 87.25% mean average (mAP), weighing only 4.98 MB improving by 6.73% 2.11% reducing parameters 2 141.02 compared YOLOV5n6 YOLOV5l6. Unlike networks that detected single or dual grades, this offered refined advantages both size, making it suitable embedded devices limited resources. Thus, the YOLOV5n6_MobileNetV3 enhanced supported intelligent harvesting research technology.

Language: Английский

---

Hyperspectral Imaging and Machine Learning for Diagnosing Rice Bacterial Blight Symptoms Caused by Xanthomonas oryzae pv. oryzae, Pantoea ananatis and Enterobacter asburiae

Plants, Journal Year: 2025, Volume and Issue: 14(5), P. 733 - 733

Published: Feb. 27, 2025

In rice, infections caused by Pantoea ananatis or Enterobacter asburiae closely resemble the bacterial blight induced Xanthomonas oryzae pv. oryzae, yet they differ in drug resistance and management strategies. This study explores potential of combining hyperspectral imaging (HSI) with machine learning for rapid accurate detection rice symptoms various pathogens. One-dimensional convolutional neural networks (1DCNNs) were employed to construct a classification model, integrating spectral preprocessing techniques feature selection algorithms comparison. To enhance model robustness mitigate overfitting due limited samples, generative adversarial (GANs) utilized augment dataset. The results indicated that 1DCNN after using uninformative variable elimination (UVE), achieved an accuracy 86.11% F1 score 0.8625 on five-class However, dominance mixed samples negatively impacted performance. After removing mixed-infection attained 97.06% 0.9703 four-class dataset, demonstrating high across different pathogen-induced infections. Key bands identified at 420–490 nm, 610–670 780–850 910–940 facilitating pathogen differentiation. presents precise, non-destructive approach plant disease detection, offering valuable insights into prevention precision agriculture.

Language: Английский

---

Optimization of Hyperparameters for SVM Classification of Citrus Diseases Using Grid Search and Cross-Validation

Lecture notes in electrical engineering, Journal Year: 2025, Volume and Issue: unknown, P. 489 - 497

Published: Jan. 1, 2025

Language: Английский

---