Convolutional Neural Networks with Fuzzy-Based Modelling: A Framework for Disease Detection in Cocoa Crops

Context: The increasing need for accurate and interpretable computer vision systems in agriculture has driven research into hybrid intelligent models. Cocoa production in Nigeria suffers significant yield losses due to diseases such as Black Pod and Cocoa Swollen Shoot Virus Disease (CSSVD), while manual inspection remains slow, subjective, and inefficient. Integrating fuzzy logic with convolutional neural networks (CNNs) offers a pathway to address diagnostic uncertainty and environmental variability. Objectives: This study aims to develop and evaluate a Fuzzy-based Convolutional Neural Network (CNN–Fuzzy) framework for the automated detection and classification of major cocoa diseases. The goal is to enhance diagnostic accuracy, improve handling of uncertain field conditions, and provide a practical decision-support tool for farmers. Methods: A hybrid architecture combining CNN-based image feature extraction with fuzzy inference rules was designed. A dataset of 12,000 annotated cocoa leaf and pod images sourced from the Cocoa Research Institute of Nigeria and open repositories (e.g., Kaggle) was used to train and validate the model. Comparative experiments were conducted against traditional machine learning classifiers and transfer learning-based models. The final framework was deployed as a mobile application optimized for offline use to support field-based disease diagnosis. Results: The proposed CNN–Fuzzy model achieved a classification accuracy of 99.99%, surpassing traditional machine learning models (75.48–80.34%) and transfer learning approaches (up to 97.27%). Field-oriented deployment demonstrated its capability to identify diseases including Black Pod and CSSVD in real time and provide context-specific remedy recommendations. Conclusion: The integration of fuzzy reasoning with deep learning significantly enhances the reliability and interpretability of cocoa disease diagnosis. The developed system offers a scalable, low-cost precision agriculture tool capable of reducing crop losses by an estimated 30–50%. This research advances digital transformation in cocoa production and lays a foundation for sustainable, AI-driven agricultural innovation in developing economies