TY  - UNPB
A1  - Hidayat, Ahmad
PB  - Fakultas Teknik Universitas Sultan Ageng Tirtayasa
ID  - eprintuntirta48674
Y1  - 2025/05/11/
UR  - https://eprints.untirta.ac.id/48674/
TI  - ANALISIS PERFORMA YOLO PADA DETEKSI 
SAMPAH ORGANIK
M1  - sarjana
N2  - YOLO (You Only Look Once) is an object detection method known for its high accuracy. This study analyzes the impact of data augmentation and object background on the performance of organic waste detection (vegetables and fruits) using YOLOv8. Two models were compared, Model 1 (without augmentation) and Model 2 (with augmentation), tested using datasets with plain and complex backgrounds. The results show that Model 2 outperformed Model 1, achieving an accuracy of 98.0% on plain backgrounds and 78.0% on complex backgrounds, compared to Model 1 96.7% and 64.0%, respectively. On average, Model 2 recorded an accuracy of 88.0%, higher than Model 1 80.3%. Additionally, Model 2 also achieved better precision, recall, and F1-score values, demonstrating that data augmentation can enhance YOLO performance in detecting organic waste. The background of the object significantly affects the model?s accuracy, especially with complex backgrounds that make it difficult for the model to distinguish the main object from other elements in the image.


Keywords: Performance Analysis, YOLO, Data Augmentation, Object Backgrounds, Organic Waste.
EP  - 97
AV  - restricted
ER  -