Enhancing Agricultural Productivity through Automated Detection of Coffee Leaf Nutrient Deficiencies

The timely identification of nutrient deficiencies in coffee plants is critical for maintaining crop health and maximizing agricultural productivity, yet traditional manual inspection methods are often slow, costly, and prone to error. This research addresses this challenge by developing an automated system for the real-time detection of nutrient deficiencies in coffee leaves, leveraging advancements in deep learning. The study implements the YOLOv11 object detection model, trained for over 50 epochs on the CoLeaf dataset, which comprises images of healthy leaves and nine distinct nutrient deficiency categories. To enhance model robustness and generalization, the images underwent comprehensive preprocessing, including isolated object extraction, auto-orientation, and contrast adjustment, followed by extensive data augmentation using techniques such as horizontal flipping, 90° rotation, blurring, and noise injection. The trained model demonstrated strong learning capabilities, achieving a stable mean Average Precision ([email protected]) exceeding 0.65, with both precision and recall stabilizing around 62%. The model showed high accuracy in identifying Potassium and Calcium deficiencies, but struggled with the underrepresented Healthy class due to significant class imbalance in the dataset. These findings confirm the viability of the YOLOv11 model for developing a practical, real-time monitoring tool for precision agriculture. The study highlights the critical need to address data imbalance through targeted augmentation or dataset expansion to further improve detection accuracy and create a more robust system for enhancing coffee cultivation.