RABBANI, NABIL RAIHAN (2026) IMPLEMENTASI MODEL DEEP LEARNING YOLOV11 UNTUK ANALISIS DAN KOREKSI POSISI DUDUK REAL-TIME. S1 thesis, Universitas Mercu Buana Jakarta.
|
Text (HAL COVER)
Cover.pdf Download (436kB) | Preview |
|
![[img]](https://repository.mercubuana.ac.id/style/images/fileicons/text.png) |
Text (BAB I)
Bab_1.pdf Restricted to Registered users only Download (44kB) |
|
|
Text (BAB II)
Bab 2.pdf Restricted to Registered users only Download (368kB) |
|
|
Text (BAB III)
Bab_3.pdf Restricted to Registered users only Download (199kB) |
|
|
Text (BAB IV)
Bab_4.pdf Restricted to Registered users only Download (764kB) |
|
|
Text (BAB V)
Bab_5.pdf Restricted to Registered users only Download (27kB) |
|
|
Text (DAFTAR PUSTAKA)
Daftar_Pustaka.pdf Restricted to Registered users only Download (152kB) |
|
|
Text (LAMPIRAN)
Lampiran.pdf Restricted to Registered users only Download (728kB) |
Abstract
This study aims to develop a Computer Vision-based system capable of detecting and classifying sitting posture in real time using the YOLOv11 (You Only Look Once Version 11) deep learning algorithm. Maintaining an improper sitting posture for extended periods may increase the risk of musculoskeletal disorders and reduce comfort during work or study activities. To address this issue, an object detection model was implemented to recognize sitting posture and classify it into two categories: Good Posture and Bad Posture. The proposed model was trained using a secondary dataset consisting of 5,036 images, which had undergone preprocessing, data augmentation, and dataset splitting with an 80:10:10 ratio for training, validation, and testing. Model training was performed using Ultralytics Platform, after which the trained model was integrated into a web-based application to enable real-time posture detection through a webcam. Experimental results indicate that the proposed model achieved a Precision of 93.7%, Recall of 93.3%, mAP50 of 96.5%, and mAP50-95 of 81.4%. Furthermore, evaluation using the confusion matrix demonstrated high classification performance, reaching 97% accuracy for the Bad Posture class and 93% for the Good Posture class. Real-time implementation testing also confirmed that the system was able to accurately classify various sitting posture conditions captured by the camera. These findings demonstrate that the proposed YOLOv11s-based approach provides reliable performance for web-based real-time sitting posture monitoring and has the potential to support healthier ergonomic habits through the application of Computer Vision technology. Keywords: Computer Vision, Deep Learning, YOLOv11, Sitting Posture Detection, Real-Time, Binary Classification. Penelitian ini berfokus pada pengembangan sistem berbasis penglihatan komputer untuk mendeteksi postur duduk secara real-time, dengan menggunakan algoritma deep learning YOLOv11 (You Only Look Once Version 11). Posisi duduk yang tidak ergonomis dalam jangka waktu yang lama dapat meningkatkan risiko gangguan muskuloskeletal dan menurunkan kenyamanan selama bekerja atau belajar di depan komputer. Oleh karena itu, tujuan penelitian ini adalah mengembangkan sistem yang mampu mendeteksi postur duduk secara otomatis melalui kamera dan mengklasifikasikannya ke dalam Good Posture dan Bad Posture. Dataset yang digunakan terdiri dari 5.036 citra yang telah melalui proses preprocessing, augmentasi data, dan pembagian dataset dengan rasio 80:10:10 untuk data pelatihan, validasi, dan pengujian. Model YOLOv11s dilatih menggunakan platform Ultralytics Platform dan diimplementasikan ke dalam aplikasi berbasis web untuk mendukung proses deteksi secara real-time. Hasil evaluasi menunjukkan bahwa model memperoleh nilai Precision sebesar 93,7%, Recall sebesar 93,3%, mAP50 sebesar 96,5%, dan mAP50-95 sebesar 81,4%. Selain itu, hasil confusion matrix menunjukkan tingkat klasifikasi yang tinggi pada kedua kelas dengan tingkat ketepatan mencapai 97% untuk kategori Bad Posture dan 93% untuk kategori Good Posture. Hasil pengujian sistem secara real-time menunjukkan bahwa model mampu mengklasifikasikan berbagai variasi posisi duduk sesuai dengan kondisi sebenarnya. Dengan demikian, sistem yang dibuat bisa memberikan umpan balik visual langsung kepada pengguna dan memiliki potensi untuk mendukung pembuatan lingkungan kerja yang lebih sehat dengan menggunakan teknologi kecerdasan buatan berbasis Computer Vision. Kata Kunci: Computer Vision, Deep Learning, YOLOv11, Deteksi Posisi Duduk, Real-Time, Klasifikasi Biner.
Actions (login required)
 |
View Item |
