Nutrient deficiency detection and classification in coffee leaves using deep learning models

Coffee is one of the most popular drink in the world, with billions of cups consumed daily. It faces nutritional deficiencies that can reduce yield and productivity. Early detection and treatment are crucial for addressing these issues. Essential nutrients like nitrogen, phosphorus, and potassium are crucial for coffee plant growth. A deep learning approach can help identify and address these deficiencies. This study uses VGG16, YOLOv8n-cls, and YOLO11s to detect and classify the observed deficiencies in coffee leaves. These models are trained on a CoLeaf-DB dataset with NPK deficiency. The VGG16 model achieved 99.67% training accuracy after training over 50 epochs. Strong overall performance using the YOLO11s model is demonstrated by the mean average precision at 0.5 and mean average precision from 0.5:0.95 measures, which also rise dramatically, reaching values over 90% and 77%, respectively. YOLOv8n-cls train model obtained a top-1 accuracy greater than 97%, reflecting strong classification performance, and the top-5 accuracy remains constant at 100%. This research aims to achieve a higher accuracy for detection and classification nutrition deficiency in an image using trained deep-learning models.