Implementasi Pseudo Labelling pada Dataset Deteksi Objek di Lingkungan Pejalan Kaki Berbasis Deep Learning

Khanafi, Nur Putra (2024) Implementasi Pseudo Labelling pada Dataset Deteksi Objek di Lingkungan Pejalan Kaki Berbasis Deep Learning. Project Report. [s.n.], [s.l.]. (Unpublished)

Text 05111940000020-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Smartphone dapat berperan penting bagi penyandang tunanetra dalam merekam data visual dan disampaikan dalam bentuk suara. Sistem ini dapat dirancang jika menerapkan object detection. Dalam proses percangan sistem kecerdasan buatan, dibutuhkan dataset yang besar berisi daftar citra dan anotasi tiap objek. Penganotasian dataset yang dilakukan secara manual dapat memperlambat proses perancangan sistem. Dengan metode Pseudo Labelling untuk menambahan data melalui pelabelan secara otomatis, dataset dapat dianotasikan lebih cepat sehingga membantu proses perancangan sistem dengan efisien. Perancangan model Pseudo Labelling dilakukan menggunakan 716 citra dataset lingkungan pejalan kaki yang dilatih dengan CNN menggunakan arsitektur YOLOv8. Peforma model dievaluasi menggunakan metrik mean Average Precision pada iou 0,5 (mAP50). Hasil dari model terbaik yang terpilih kemudian digunakan untuk melabel dan memberikan anotasi pada dataset kedua yang berisi 740 citra.
============================================================================================================================
Smartphone plays a crucial role for visually impaired individuals in recording visual data and presenting it in audio format. Such a system can be designed by implementing object detection. During the process of artificial intelligence system development, a large dataset containing a list of images and annotations for each object is required. Manual annotation of the dataset can slow down the system development process. By using the Pseudo Labelling method to augment data through automatic labeling, the dataset can be annotated more quickly, thus aiding in the efficient system development process. The Pseudo Labelling model design was conducted using 716 images from a pedestrian environment dataset trained with a CNN using the YOLOv8 architecture. The model's performance was evaluated using the mean Average Precision metric at an intersection over union (iou) of 0.5 (mAP50). The results from the selected best model were then used to label and annotate a second dataset containing 740 images.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Object Detection, Pseudo Labelling, YOLOv8, mAP, Object Detection, Pseudo Labelling, YOLOv8, mAP
Subjects:	T Technology > T Technology (General) > T11 Technical writing. Scientific Writing T Technology > T Technology (General) > T174 Technological forecasting T Technology > TA Engineering (General). Civil engineering (General) > TA1637 Image processing--Digital techniques T Technology > TA Engineering (General). Civil engineering (General) > TA1650 Face recognition. Optical pattern recognition.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Nur Putra Khanafi
Date Deposited:	13 Feb 2024 07:35
Last Modified:	13 Feb 2024 07:35
URI:	http://repository.its.ac.id/id/eprint/107226

Actions (login required)

View Item