Deteksi Senjata Genggam Menggunakan Faster R-Cnn Inception V2

Pradana, Ivandi Christiani (2022) Deteksi Senjata Genggam Menggunakan Faster R-Cnn Inception V2. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 07211740000041-Undergraduate_Thesis.pdf Restricted to Repository staff only Download (9MB) | Request a copy

Abstract

Senjata genggam kelas pisau sering digunakan dalam kegiatan kriminal di Indonesia. Sering kali objek pisau yang terekam sulit terlihat dengan mata telanjang. Proses deteksi senjata genggam kelas pisau bisa dibantu dengan pengolahan citra menggunakan Deep Learning. Penelitian ini bertujuan untuk mengaplikasikan konsep Deep Learning dan Tensorflow Object Detection untuk melatih model Faster R-CNN Inception V2 untuk bisa mendeteksi senjata genggam kelas pisau dalam citra digital. Dalam penelitian ini, model yang terlatih bisa menandai benda yang diduga sebagai senjata genggam kelas pisau dalam gambar dengan kotak penanda. Model yang dibuat dari penelitian ini dilatih dengan kumpulan dataset berisi gambar senjata genggam kelas pisau, dataset didapat dan terdiri dari rekaman bela diri pisau dan kumpulan pisau dengan bentuk dan warna yang beragam. Penelitian ini meneliti akurasi model Faster R-CNN Inception V2 yang dilatih dalam mendeteksi senjata genggam kelas pisau. Hasil akhir dari proses pengembangan model Faster R-CNN Inception V2 ini adalah model yang berhasil mendeteksi senjata genggam kelas pisau dengan akurasi sebanyak 87%, hasil akurasi didapatkan dari pengujian terhadap 475 gambar digital yang dilakukan di Google Colab.
==============================================================================================================================
Knife-class handheld objects are often used in criminal activities in Indonesia. Often the recorded knife object is difficult to see with the naked eye. The detection process for knife-class handheld objects can be assisted by image processing using Deep Learning. This study aims to apply the concepts of Deep Learning and Tensorflow Object Detection to train the Faster R-CNN Inception V2 model to detect knife-class handheld objects in digital images. In this study, trained models were able to mark objects suspected of being knife in images with marking boxes. The model created from this study was trained with a dataset containing images of knife-class handheld weapons, the dataset was obtained and consisted of recordings of knife self-defense and a collection of knives with various shapes and colors. This study examines the accuracy of the Faster R-CNN Inception V2 model which is trained in detecting knife-class handheld objects. The final result of the process of developing the Faster R-CNN Inception V2 model is a model that successfully detects knife-class handheld objects with an accuracy of 87%, the accuracy results are obtained from testing on 475 digital images conducted on Google Colab.

Item Type:	Thesis (Other)
Additional Information:	RSKom 006.312 Pra d-1 2022
Uncontrolled Keywords:	Inception-V2. Deep Learning. TensorFlow. Knives.
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science. EDP
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Computer Engineering > 90243-(S1) Undergraduate Thesis
Depositing User:	Mr. Marsudiyana -
Date Deposited:	15 Jun 2026 06:41
Last Modified:	15 Jun 2026 06:41
URI:	http://repository.its.ac.id/id/eprint/133812

Actions (login required)

View Item