Evaluasi Performa Algoritma YOLOv8 dalam Deteksi dan Klasifikasi Posisi Tidur

Julivano, Hernawan Yudha and -, Gunawan Ariyanto, S.T., M.Comp.Sc., Ph.D. (2024) Evaluasi Performa Algoritma YOLOv8 dalam Deteksi dan Klasifikasi Posisi Tidur. Skripsi thesis, Universitas Muhammadiyah Surakarta.

	PDF (Naskah Publikasi) Naskah Publikasi.pdf Download (4MB)
	PDF (Surat Pernyataan Publikasi) Surat Pernyataan Publikasi.pdf Restricted to Repository staff only Download (162kB)

Abstract

Object detection using computer vision technology enables computers to recognize and determine the location and type of objects in images or videos. Traditional methods such as HOG and SIFT have limitations in handling object variations, while deep learning approaches such as CNN and YOLO (You Only Look Once) algorithms demonstrate better performance. YOLO processes images in real-time and predicts bounding boxes and object classes in a single neural network. This study aims to evaluate the ef ectiveness of the YOLOv8 algorithm in detecting humans and classifying sleeping positions, such as supine, left lateral, right lateral, and others. The test results show that YOLO successfully detects humans in bed with over 90% accuracy on a similar dataset to the training data, but performance drops significantly to 25%-40% accuracy on dif erent datasets. Testing on mixed datasets still shows accuracy above 90%. In image classification, YOLO also performs well with over 85% accuracy on similar datasets, but drops to 20% on different datasets. Balanced and proportional training datasets are crucial for optimal performance. The dif erence between YOLOv8n and YOLOv8s is not significant, but YOLOv8n is more suitable for devices with low computation and small datasets, whereas YOLOv8s has better overall performance but requires higher computation and larger datasets.

Item Type:	Thesis (Skripsi)
Uncontrolled Keywords:	deteksi, klasifikasi, posisi tidur, yolov8
Subjects:	T Technology > TZ Technical Information
Divisions:	Fakultas Ilmu Komunikasi dan Informatika > Teknik Informatika
Depositing User:	HERNAWAN YUDHA JULIVANO
Date Deposited:	13 Aug 2024 02:38
Last Modified:	15 Aug 2024 03:00
URI:	http://eprints.ums.ac.id/id/eprint/126585

Actions (login required)

View Item