Deep Learning Implementation for Peruvian Blueberry Export Standards: A YOLOv8n Solution

This research introduces a computer vision system based on YOLOv8n, an efficient convolutional neural network (CNN) architecture, for real-time classification of blueberry maturity stages. The developed solution automates the detection of three distinct maturity stages (green, semi-ripe, and ripe) to support precision agriculture applications and ensure compliance with Peruvian Technical Standards (NTP) for blueberry exports. Through extensive experimentation using a curated dataset of 550 field-acquired images, the optimized model achieves competitive performance across key evaluation metrics—including precision, recall, average precision (AP), mean average precision (mAP), and F1 score—when compared to relevant previous studies. These results demonstrate the potential of the approach to enhance harvesting efficiency and guarantee adherence to international export quality requirements through automated visual inspection.