DOI: 10.7763/IJCTE.2026.V18.1398
Attention-driven Deep Convolutional Neural Network for Maize Leaf Nutrient Deficiency Recognition
2. MAEER’s MIT ACSC Alandi, Pune, Savitribai Phule Pune University, Pune, India
3. Department of Electronic Science, Fergusson College (Autonomous), Pune, India
4. Department of Electronic Science, Savitribai Phule Pune University, Pune, India
Email: shubhangigaikar11@gmail.com (G.S.); drmsz29@gmail.com (M.Z.); adshaligram@gmail.com (A.S.); kailas.sapnar@fergusson.edu (K.S.)
*Corresponding author
Manuscript received October 26, 2025; revised February 3, 2026; accepted April 9, 2026; published July 17, 2026
Abstract—This research focuses on developing an Attention-Driven Deep Convolutional Neural Network (ADCNN) to predict nutrient deficiencies in maize using an image dataset. The main advantage of the ADCNN model is its use of attention mechanisms, which enable it to focus on relevant features at the leaf level. This results in more accurate and transparent classification compared to previous methods that relied solely on Convolutional Neural Networks (CNNs). The ADCNN was trained and tested on a dataset covering six nutrient classes: healthy (ALL Present), general nutrient imbalance (ALLAB), nitrogen (NAB), phosphorus (PAB), potassium (KAB), and zinc (ZNAB). Its performance was compared with three deep learning models: the baseline CNN, Visual Geometry Group 19 (VGG19), and hybrid Convolutional Neural Network-Support Vector Machine (CNN-SVM). Results indicate that the ADCNN performs better than any of the competing models, with accuracy, precision, recall, and F1-scores of nearly 99.48%, which demonstrates its efficacy. The attention mechanism improves the model’s discriminative capability, enabling it to identify signs of deficiencies in appearance. The ADCNN has enhanced generalization, stability, and response time, which are superior to the conventional CNN architecture. This research shows that using a deep attention process enables real-time or automated plant health monitoring based on nutrient status. The ADCNN model is a stable, efficient system that can be used to create intelligent agricultural systems that enhance sustainable agricultural practices by optimizing nutrient use. Future research should focus on collecting real-world data across various environmental conditions and on integrating Internet of Things (IoT)-enabled platforms to advance multimodal data collection for a more comprehensive plant health assessment.
Keywords—maize leaf, nutrient deficiency detection, attention mechanism, deep learning, Convolutional Neural Network (CNN), precision agriculture, image classification
Cite: Gaikar Shubhangi, Madhukar Zambare, Arvind Shaligram, and Kailash Sapnar, " Attention-driven Deep Convolutional Neural Network for Maize Leaf Nutrient Deficiency Recognition," International Journal of Computer Theory and Engineering, vol. 18, no. 3, pp. 164-179, 2026.
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).