Statistical Methods for Enhancing Performance in Precision Agriculture

Agriculture is key to securing food for the 9.7 billion people projected to inhabit the world by 2050. However, managing plant health remains a significant challenge for crop productivity. Traditional approaches for identifying plant species, as well as for disease classification and weed detection, are laborious, inefficient, and time-consuming. We use deep learning architecture and transfer learning methods to provide automated, rapid, proven, and accurate plant solutions. Plant Leaf Classification uses deep learning models to recognize species of plants from their leaves, aiding in medical, biodiversity, and ecology research. Transfer learning plays a vital role in leaf disease classification, detecting diseases as quickly as possible ,and taking timely action. Object Detection Models help in Leaf Disease Detection and Weed Detection in real-time, when used in a way that allows for controlled application, rather than spraying pesticides throughout the entire field. ResNet50, DenseNet201, YOLOv7, and YOLOv8 are evaluated in the study based on performance metrics such as accuracy and loss curves, class distribution analysis, and confidence curves. Model efficiency is briefly addressed using key metrics: precision, recall, f1 score, and MAP (Mean Average Precision). These findings inform the best model and hyperparameter selection for plant classification tasks. More specifically, this study enhances the accuracy and sustainability of plant management, advancing precision agriculture toward greater efficiency and data-driven solutions