Vision-Based Chili Quality Assessment Using YOLO

Lee, Qing Hwee (2026) Vision-Based Chili Quality Assessment Using YOLO. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text LEE QING HWEE_Full Text.pdf Restricted to Registered users only Download (2MB)

Abstract

Manual quality assessment of chilis is often inconsistent due to subjective human judgment, creating a need for objective, automated solutions in precision agriculture. This project aims to develop an automated vision system to classify chili based on size and colour features while identifying the most efficient YOLO model for the task. The methodology involved a hardware setup utilizing a lighting box equipped with a USB webcam, LED desk lamp and a standard computer to ensure consistent data capture for accuracy. The system was developed using an annotated dataset processed through automated augmentation and comparative training across YOLOv8, YOLOv11, YOLOv12, and YOLO26. Results demonstrated that YOLOv11 is the most stable model, providing consistent performance for budget-friendly inspection and classification to prioritize red chili selection. Ultimately, the system offers a reliable, low-cost replacement for manual inspection that improves post-harvest efficiency and supports sustainable agricultural infrastructure