ANALISIS KOMPARATIF KINERJA RF-DETR DAN YOLOV11 UNTUK DETEKSI DAN KLASIFIKASI KUALITAS CABAI

Abstract

Increasing demands for the quality and consistency of horticultural commodities highlight the limitations of manual
sorting methods, which tend to be subjective and inefficient. This study aims to develop and evaluate an automated
system for detecting and classifying chili pepper quality using a deep learning-based computer vision approach. Two
state-of-the-art object detection models, namely YOLOv11 and RF-DETR, were implemented and compared using a
dataset of chili pepper images classified into fresh, unripe, and damaged categories. The research methodology
employed a quantitative experimental approach encompassing data collection, instance segmentation annotation,
preprocessing, model training, and performance evaluation. The results show that both models demonstrate good
detection capabilities, with YOLOv11 excelling in inference speed, making it suitable for real-time applications, while
RF-DETR exhibits higher detection accuracy and stability under complex visual conditions. These findings indicate
a balance between computational efficiency and detection accuracy. This research contributes to the development of
computer vision-based automated systems to improve the efficiency and objectivity of the agricultural product grading
process.