Revolutionizing Agriculture: Machine and Deep Learning Solutions for Enhanced Crop Quality.

Advancements in technology are revolutionizing agriculture by not only increasing yield per hectare but also improving crop quality and ensuring environmental sustainability. Weeds remain a major challenge as they compete with crops for vital resources. Traditional weed control methods, like uniform spraying, are both costly and environmentally harmful. To address this, the study employs Machine Learning (ML) and Deep Learning (DL) techniques for weed detection and classification. ML models use extracted features such as Hu moments, GLCM, LBP, and entropy from the CottonWeedID15 and Early Crop-Weed datasets. YOLOv8m is used for detection, while classifiers like SVM, Random Forest, and ANN—enhanced with SMOTE—handle classification. SVM with a polynomial kernel achieves 99.5% accuracy on EarlyCrop-Weed, and ANN achieves 89% on CottonWeedID15. DL models including VGG16, DenseNet, Xception, and ConvNeXtBase are trained on raw, balanced datasets. Among them, ConvNeXt paired with Random Forest delivers the best performance—98% on EarlyCrop-Weed and 90% on CottonWeedID15. These results demonstrate the effectiveness of ML and DL in providing precise, efficient, and eco-friendly weed management solutions. Keywords: Weed Detection, Machine Learning, Deep Learning, YOLOv8m, Feature Extraction, ConvNeXt.