Evaluation of a Convolutional Neural Network-based Model for Accurate Detection and Multi-class Classification of Economically Significant Soybean Pests using RGB Image Data

Soybean (Glycine max) is a vital legume crop cultivated worldwide for its high protein and oil content. However, its productivity is significantly affected by various insect pests that damage leaves, stems and pods during different growth stages. Pests like aphids, armyworms and bollworms can cause substantial yield losses if not detected and controlled promptly. Manual pest monitoring through field scouting remains the dominant practice but is labour-intensive, inconsistent and prone to human error, particularly in large-scale farming systems. The integration of artificial intelligence, particularly deep learning, offers a scalable solution for automating pest detection.

This study presents a convolutional neural network (CNN) model designed to detect and classify seven major insect pests commonly found in soybean crops. A dataset consisting of 2,450 RGB images was collected from online resources. The images were preprocessed, augmented and split into training, validation and test sets. The CNN was built using six convolutional layers with ReLU activations, followed by max-pooling layers, a fully connected dense layer and a softmax output layer for classification. The model was trained for 100 epochs using the Adam optimizer and sparse categorical cross-entropy loss.

The final model achieved an overall accuracy of 96.88% on the test set. Class-wise evaluation showed high classification matrices across all categories. ROC-AUC values reached 1.00 for all pest classes, indicating excellent classification performance. Precision-recall curves also showed high average precision scores, confirming the model’s reliability. These results demonstrate the model’s potential for practical application in real-time pest monitoring systems used in precision agriculture.