Pembelajaran pergerakan shuttlecock pada permainan badminton menggunakan yolov4

Alfando, Mario (2022) Pembelajaran pergerakan shuttlecock pada permainan badminton menggunakan yolov4. Bachelor thesis, Universitas Pelita Harapan.

Preview	Text (Tittle) Tittle.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (154kB) | Preview
Preview	Text (Abstract) Abstract.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (185kB) | Preview
Preview	Text (ToC) ToCpdf.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (868kB) | Preview
	Text (Chapter1) Chapter1.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (752kB)
	Text (Chapter2) Chapter2.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (2MB)
	Text (Chapter3) Chapter3.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (2MB)
	Text (Chapter4) Chapter4.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (5MB)
	Text (Chapter5) Chapter5.pdf Restricted to Registered users only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (430kB)
Preview	Text (Bibliography) Bibliography.pdf Available under License Creative Commons Attribution Non-commercial Share Alike. Download (604kB) | Preview
	Text (Appendices) Appendices.pdf Restricted to Repository staff only Available under License Creative Commons Attribution Non-commercial Share Alike. Download (2MB)

Abstract

The rise of Sports Analytics has been applied in several popular sports in the world. Badminton, which is one of the most popular sports in Indonesia, did not take part in this research, aka very little research in sports analytics to improve performance. Sports analytics research this time will study the shuttlecock movement in badminton games. After all the data needed to carry out the data training process is available, a cloud-based training data training process will be carried out using a notebook from google collaboratory pro which has provided a Tesla P100PCI-EE-16GB GPU, CUDA version 11010, CUDNN version 7.6.5, CUDNN_Haf , and Opencv with version 3.2.0. The time used in the data training process runs for 5 hours 3 minutes. After the training data is successful, it will produce output in the form of a weights file that can be used as a benchmark in detecting. The weights files are yolov4_1000.weights, yolov4_2000.weights, yolov4_3000.weights, yolov4_4000.weights, yolov4_best.weights, yolov4_final.weights, yolov4_last.weights. Testing the model on a video that has 7 seconds duration got 171 times the shuttlecock was detected with an AP value of 95.9%. When the shuttlecock is at a high speed or greater than 253 km/h and has a high percentage of accuracy value, yolov4 also has difficulty in drawing bounding boxes because the distance from frame to frame moves so far. Has a different result When the shuttlecock is at a speed below 50km/h and has a low percentage value yolov4 can accurately describe the bounding boxes. / Maraknya Sport Analytics yang sudah diterapkan dalam beberapa cabang olahraga populer dunia. Badminton merupakan salah satu olahraga paling populer di Indonesia tidak mendapat bagian dalam penelitian ini alias sangat minim penelitian dalam sport analytics guna untuk meningkatkan prestasi. Penelitian sports analytics kali ini akan melakukan pembelajaran terhadap pergerakan shuttlecock dalam permainan badminton. Setelah semua data yang diperlukan untuk melakukan proses training data sudah tersedia, maka akan dilakukan proses training data yang berbasis cloud training menggunakan notebook dari google colaboratory pro yang telah menyediakan GPU Tesla P100-PCIE-16GB, CUDA versi 11010, CUDNN versi 7.6.5, CUDNN_Half , dan Opencv dengan versi 3.2.0. Waktu yang digunakan dalam proses training data berjalan selama 5 jam 3 menit. Setelah training data berhasil maka akan menghasilkan keluaran berupa weights file yang dapat digunakan sebagai patokan dalam melakukan deteksi. Weights file tersebut yaitu yolov4_1000.weights,yolov4_2000.weights,yolov4_3000.weights, yolov4_4000.weights, yolov4_best.weights, yolov4_final.weights, yolov4_last.weights. Pengujian model terhadap video dengan durasi 7 detik mendapatkan hasil sebanyak 171 kali shuttlecock terdeteksi dengan nilai AP 95,9%. Ketika shuttlecock berada dalam kecepatan tinggi atau lebih besar dari 253 km/jam dan memiliki nilai persentase akurasi yang tinggi sekalian yolov4 juga mendapatkan kesulitan dalam menggambarkan bounding boxes karena pergerakan jarak dari frame ke frame yang begitu jauh. Memiliki hasil yang berbeda Ketika shuttlecock berada dalam kecepatan di bawah 50 km/jam dan memiliki nilai persentase rendah yolov4 mampu menggambarkan bounding boxes dengan tepat.

Actions (login required)

View Item