Edge AI-Based Defect Detection in White Pepper Using CLAHE-Based Pre-processing and YOLO

Defect detection in white pepper plays a critical role in maintaining export quality, particularly for Indonesia as a leading producer. This study presents a real-time classification system using the YOLOv8 deep learning model, deployed on the NVIDIA Jetson Orin Nano device. To improve image quality under varying lighting conditions, CLAHE image enhancement is applied prior to detection. The system classifies white pepper into normal and defective categories with high accuracy and low latency. A Streamlit-based interface is integrated to provide interactive real-time visualization on edge devices, eliminating the need for internet connectivity.