Choirunnisa Zalfaa Nabilah, . (2026) PENGEMBANGAN SISTEM DETEKSI GERAKAN HOME WORKOUT SECARA REAL-TIME BERBASIS COMPUTER VISION UNTUK KOREKSI POSTUR DAN PERHITUNGAN REPETISI. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.
![[img]](http://repository.upnvj.ac.id/style/images/fileicons/text.png) |
Text
ABSTRAK.pdf Download (174kB) |
|
Text
AWAL.pdf Download (942kB) |
|
Text
BAB 1.pdf Restricted to Repository UPNVJ Only Download (106kB) |
|
Text
BAB 2.pdf Restricted to Repository UPNVJ Only Download (833kB) |
|
Text
BAB 3.pdf Restricted to Repository UPNVJ Only Download (1MB) |
|
Text
BAB 4.pdf Restricted to Repository UPNVJ Only Download (950kB) |
|
Text
BAB 5.pdf Download (22kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (163kB) |
|
Text
RIWAYAT HIDUP.pdf Restricted to Repository staff only Download (94kB) |
|
Text
LAMPIRAN.pdf Restricted to Repository UPNVJ Only Download (1MB) |
|
Text
HASIL PLAGIARISME.pdf Restricted to Repository staff only Download (29MB) |
|
Text
ARTIKEL KI.pdf Restricted to Repository staff only Download (400kB) |
Abstract
Independent home workouts are not always accompanied by posture monitoring and repetition counting, allowing movement errors to occur without feedback. This study developed a computer-vision-based desktop application to detect and analyze five home workout exercises in real time in single-person and multi-person modes. The study used a quantitative approach and an adaptation of the Rapid Application Development method. Data were obtained from primary recordings and the Kinetics-700 dataset, totaling 1,745 clips after preprocessing and covering jumping jacks, squats, mountain climbers, knee push-ups, and sit-ups. The system integrates YOLO for user detection and tracking, MediaPipe Pose for landmark extraction, LSTM-Attention for exercise classification, and kinematic rules for phase, posture, and repetition analysis. The model achieved 97.42% accuracy on 349 test samples, with macro-averaged precision, recall, and F1-score of 97% each. After integration, correct-prediction rates reached 92% in single-person mode and 90% in multi-person mode, with mean end-to-end latencies of 56.919 ms and 74.880 ms, respectively. All 35 functional scenarios and 17 posture-validation scenarios produced the expected outputs. Pose estimation remained stable under the tested variations, but identity tracking was inconsistent when participants crossed paths or left the frame.
| Item Type: | Thesis (Skripsi) |
|---|---|
| Additional Information: | [No.Panggil: 2210511070] [Pembimbing 1: Supriyanto] [Pembimbing 2: Teguh Firmansyah] [Penguji 1: Didit Widiyanto] [Penguji 2: I Wayan Rangga Pinastawa] |
| Uncontrolled Keywords: | computer vision, home workout, LSTM-Attention, posture correction, repetition counting |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software |
| Divisions: | Fakultas Ilmu Komputer > Program Studi Informatika (S1) |
| Depositing User: | CHOIRUNNISA ZALFAA NABILAH |
| Date Deposited: | 08 Sep 2026 07:37 |
| Last Modified: | 08 Sep 2026 07:37 |
| URI: | http://repository.upnvj.ac.id/id/eprint/52247 |
Actions (login required)
 |
View Item |