Editorial: Harnessing machine learning to decode plant-microbiome dynamics for sustainable agriculture

increasingly turning to machine learning, a subset of artificial intelligence that enables computers to learn from data and make predictions (De Souza et al., 2020). Deep-learning models, a powerful type of machine learning, are particularly effective for analyzing complex biological data. These models are built from layers of interconnected nodes that process input data, such as microbial DNA sequences or plant images, to identify patterns and relationships. Developers must make critical decisions when designing these models, such as choosing the number and type of layers, selecting the data features to focus on (e.g., specific microbial traits), and determining how the model learns from errors (Zhou and Gallins, 2019). These choices depend on the specific problem, such as detecting crop diseases or predicting yield, and are guided by the need for accuracy, computational efficiency, and applicability to real-world farming conditions (Zhou and Gallins, 2019).The development of a machine vision-based method using an enhanced YOLOv5s model for grading individual peanut pod rot, which is a major plant disease affecting peanut production were investigated in a recent paper published by Liu et al. (2024b). YOLO is a real-time object detection algorithm known for its speed and efficiency. Unlike traditional methods that repurpose classifiers or localizers to perform detection, YOLO frames object detection as a single regression problem, directly predicting bounding boxes and class probabilities from full images in one evaluation. This model, which relies on deep-learning principles to process images, incorporates a Shuffle Attention module to focus on key visual features and replaces the loss function CIoU with EIoU to improve accuracy in distinguishing non-rotted and rotten peanuts in complex backgrounds. The study also highlighted the potential for future research to enhance prediction performance for different peanut varieties and to consider factors like rotten kernel rate for better yield estimation (Liu et al., 2024b). In another study by V. et al., (2024), the possibility of using a machine vision-based approach for grading individual peanut pod rot using an improved YOLOv5s algorithm were investigated. The study addresses the challenges of visually identifying and classifying peanut pod rot by introducing a Shuffle Attention module to enhance feature representation and accuracy in complex backgrounds. The proposed model demonstrated high recognition rates for non-rotted and rotten peanuts, offering a promising solution for automated grading of peanut pod rot, providing advancements in disease resistance evaluation and germplasm selection in peanut breeding (V. et al., 2024). Another use of YOLO algorithms was reported by Wang et al. (2024b) Plant diseases pose a significant threat to global agriculture by negatively impacting crop yield and quality (Yoosefzadeh Najafabadi, 2021). Despite the challenges associated with identifying and classifying these diseases, a new approach leveraging deep learning algorithms and convolutional neural networks (CNNs) has been proposed to accurately detect and categorize leaf diseases in economically important crops such as strawberries, peaches, cherries, and soybeans (Prince et al., 2024). For this aim, a research focuses on categorizing 10 disease classes for these crops, comprising 6 diseased classes and 4 healthy classes, using a CNN-support vector machine (SVM) model (Prince et al., 2024). Various pre-trained models were employed, with the proposed model achieving an average accuracy of 99.09%, outperforming established models like VGG16. The model utilizes Class Activation Maps generated through the Grad-CAM technique to visually illustrate detected diseases and produce heatmaps highlighting the areas requiring classification (Prince et al., 2024). The FCHF-DETR model developed by Xin and Li (2024), an enhancement of RT-DETR-R18, addressed the challenges of detecting tomato leaf diseases with FasterNet, Cascaded Group Attention, and HSFPN. Using a dataset of 3147 images, the model achieved high precision and recall while reducing computational demands. In addressing the challenge of identifying tea plant diseases amidst complex backgrounds, the ECA-ResNet50 model improved the ResNet50 architecture by using a multilayer small convolution kernel strategy and introducing the ECA attention mechanism (Li and Zhao, 2025). convolutions for feature extraction and incorporates a transformer encoder with cross-attention for global perspective refinement. This approach improves classification performance on hyperspectral corn image datasets, demonstrating its effectiveness over current methods (Wang et al., 2025).

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.