Nutrient deficiency in coffee plants significantly impacts yield and quality, necessitating accurate and timely detection for effective agricultural management. This study proposes a deep learning-based approach for detecting and classifying nutrient deficiencies in coffee leaves using advanced convolutional neural network (CNN) models. We evaluate the performance of proposed models, including LeafNet, EfficientNet, VGG19, ResNet50, and MobileNetV2. The models are trained and tested on the CoLeaf-Augmented dataset, a comprehensive collection of coffee leaf images showcasing various nutrient deficiency symptoms, accessible via Kaggle CoLeaf-Augmented Dataset. Among the proposed models, MobileNetV2 demonstrates superior performance, achieving the highest classification accuracy due to its lightweight architecture and efficiency in feature extraction. The results highlight MobileNetV2’s potential for real-time, resource-constrained applications in precision agriculture, outperforming other models in both accuracy and computational efficiency. This study underscores the effectiveness of tailored deep learning models for nutrient deficiency detection, offering a scalable solution for coffee crop health monitoring.

Keywords: - Nutrient deficiency detection , Coffee leaves , Deep learning, Convolutional neural networks (CNN), MobileNetV2,LeafNet, EfficientNet,VGG19, ResNet50,CoLeaf-Augmented dataset,Precision agriculture, Classification accuracy