Implementasi machine learning untuk tracking dan menghitung orang menggunakan SSD mobilenet dan YOLO = Implementation of machine learning for people tracking and counting using SSD mobilenet and YOLO

Hermawan, Timothius (2021) Implementasi machine learning untuk tracking dan menghitung orang menggunakan SSD mobilenet dan YOLO = Implementation of machine learning for people tracking and counting using SSD mobilenet and YOLO. Bachelor thesis, Universitas Pelita Harapan.

[img] Text (Title)
Title - 2021-03-03T100409.826.pdf
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (710kB)
[img]
Preview
Text (Abstract)
Abstract.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (166kB) | Preview
[img]
Preview
Text (ToC)
ToC.pdf
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (194kB) | Preview
[img]
Preview
Text (Chapter 1)
Chapter 1.pdf
Available under License Creative Commons Attribution Share Alike.

Download (165kB) | Preview
[img] Text (Chapter 2)
Chapter 2.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Share Alike.

Download (498kB)
[img] Text (Chapter 3)
Chapter 3.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Share Alike.

Download (144kB)
[img] Text (Chapter 4)
Chapter 4.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Share Alike.

Download (1MB)
[img] Text (Chapter 5)
Chapter 5.pdf
Restricted to Registered users only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (98kB)
[img]
Preview
Text (Bibliography)
Bibliography.pdf
Available under License Creative Commons Attribution Share Alike.

Download (164kB) | Preview
[img] Text (Appendices)
Appendices.pdf
Restricted to Repository staff only
Available under License Creative Commons Attribution Non-commercial Share Alike.

Download (1MB)

Abstract

Pada tahun 2020, pandemi virus corona membuat social distancing diterapkan di seluruh dunia. Upaya ini dilakukan untuk mengurangi cepatnya penyebaran virus. Di Indonesia, salah satu upaya pemerintah untuk mendisiplinkan social distancing adalah dengan mengurangi kapasitas pengunjung di tempat-tempat umum seperti mall, rumah makan, dan lain-lain agar lebih banyak ruang gerak dan mengurangi kontak fisik satu sama lain. Untuk membantu upaya pemerintah, machine learning bisa diimplementasikan untuk menghitung orang yang masuk dan keluar. Dengan input sebuah video dari kamera, mesin bisa mengenali objek di dalam gambar dan mengikuti arah pergerakannya. Arah pergerakan tersebut bisa digunakan untuk menghitung jumlah orang yang masuk dan keluar dari ruangan. Untuk mendeteksi objek, ada 2 algoritma yang cukup populer digunakan yaitu You Only Look Once (YOLO) dan Single Shot Detection (SSD) MobileNet. Berdasarkan percobaan, SSD bisa mendeteksi objek berukuran kecil pada gambar (lebih kecil dari grid yang digunakan untuk deteksi) sedangkan YOLO kesulitan untuk mendeteksi objek yang berukuran relatif kecil. Maka SSD memiliki akurasi yang lebih tinggi dibandingkan dengan YOLO. Program untuk people detection and tracking dijalankan oleh Raspberry Pi 3B+ sebagai embedded device. Raspberry Pi dapat mendeteksi video dengan kecepatan deteksi berkisar antara 0.55- 1 detik atau memiliki kecepatan maksimal 2 fps menggunakan program ini. Dengan menggunakan model SSD dan training data yang dianotasi sendiri berisikan orang dan kerumunan yang diambil dari atas, model memiliki precision sebesar 31%, recall sebesar 27%, dan akurasi sebesar 28%. Ketika diuji menggunakan video pengujian akurasi model ini bisa mencapai 100%. Embedded device ini bisa dikembangkan kemudian untuk digabung ke dalam sistem yang lebih besar./In 2020, the coronavirus pandemic made social distancing implemented around the world. This effort was made to reduce the rapid spread of the virus. In Indonesia, one of the government's efforts to discipline social distancing is to reduce the capacity of visitors in public places such as malls, restaurants, and others. To aid government efforts, machine learning could be implemented to count people entering and leaving. With the input of a video from the camera, the machine can recognize objects in the image and follow the direction of its movement. The direction of movement can be used to count the number of people entering and leaving the room. To detect objects, there are 2 algorithms that are quite popular to use, namely You Only Look Once (YOLO) and MobileNet Single Shot Detection (SSD). Based on experiments, SSDs can detect small objects in images (smaller than grid size use for detecting object), while YOLO has difficulty detecting relatively small objects. So SSD has higher accuracy compared to YOLO. The program for people detection and tracking is run by the raspberry pi 3B + as an embedded device. Raspberry Pi can detect videos with speeds ranging from 0.55-1 second or has a maximum speed of 2 fps using this program. By using the SSD model and self-annotated training data containing people and crowds taken from above, the model has a precision of 31%, a recall of 27%, and an accuracy of 28%. When tested using a testing video, the accuracy of this model can reach 100%. This embedded device can be developed to be integrated into a larger system in the future.

Item Type: Thesis (Bachelor)
Creators:
CreatorsNIMEmail
Hermawan, TimothiusNIM01032170010timothiushermawan58@gmail.com
Contributors:
ContributionContributorsNIDN/NIDKEmail
Thesis advisorMartoyo, IhanNIDN0318057301ihan.martoyo@uph.edu
Thesis advisorPrasetya, KusnoNIDN0718087903kusno.prasetya@uph.edu
Uncontrolled Keywords: machine learning; image processing; people tracking and counting; you only look once (YOLO); single shot detection (SSD) mobilenet.
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions: University Subject > Current > Faculty/School - UPH Karawaci > Faculty of Science and Technology > Electrical Engineering
Current > Faculty/School - UPH Karawaci > Faculty of Science and Technology > Electrical Engineering
Depositing User: Users 3918 not found.
Date Deposited: 03 Mar 2021 03:08
Last Modified: 03 Mar 2021 03:08
URI: http://repository.uph.edu/id/eprint/26369

Actions (login required)

View Item View Item