Coffee disease classification using Convolutional Neural Network based on feature concatenation

Abstract

Coffee is a significant global agricultural commodity, and improving its production and maintaining quality is crucial. However, coffee plants are susceptible to various diseases that can lower production and quality. Early detection and identification of these diseases are critical in overcoming these challenges. In this study, we propose a deep learning approach for the identification and classification of coffee diseases using Convolutional Neural Networks (CNNs). Our research is divided into three phases: image preprocessing, feature extraction, and classification. Gaussian filtering and data augmentation techniques were applied to enhance the robustness of the model and reduce noise. We used a CNN to extract high-level features by combining GoogLeNet-based and RESNET-based architecture, which can capture more complex and meaningful characteristics of the input images, such as shapes, objects, and patterns, and are important for tasks such as object recognition and classification. The extracted features were then classified using Multi-Layer Perceptrons (MLPs), machine learning, and ensemble classifiers. Our proposed model achieved a testing accuracy of 99.08%, outperforming other classifiers. The results indicate that proper image preprocessing, data augmentation, and CNN provide an efficient classification method for identifying and classifying coffee diseases.

Keywords

• Coffee disease
• Convolutional Neural Network
• Feature concatenation
• Machine learning
• Segmentation