Resource-Aware Framework untuk Deteksi Objek pada Video Streaming Multi-Kamera

Romadhona, Yasinta (2024) Resource-Aware Framework untuk Deteksi Objek pada Video Streaming Multi-Kamera. Masters thesis, Institut Teknologi Sepuluh Nopember.

Text 6025221027-Master_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (4MB) | Request a copy

Abstract

Video stream mining memerlukan banyak resource karena membutuhkan kemampuan untuk mengeksplorasi dan mengekstraksi informasi penting dari aliran data video secara terus menerus. Keadaan ini dapat mengakibatkan resource menjadi terbatas, yang menyebabkan sistem bekerja kurang efektif. Salah satu cara untuk memenuhi kebutuhan ini adalah dengan mengalokasikan lebih banyak resource, namun hal ini dapat meningkatkan biaya.
Penelitian ini mengembangkan resource-aware framework untuk deteksi objek dalam lingkungan video streaming multi-kamera. Pengaturan Algorithm Granularity Settings (AGS) digunakan untuk melakukan adaptasi penggunaan resource pada berbagai komponen sistem (CPU, RAM, dan disk storage) guna menjaga stabilitas dan efisiensi kinerja sistem. Keandalan kerangka kerja ini diuji untuk deteksi manusia menggunakan algoritma deep learning (YOLOv7-tiny) dengan input dari video streaming multi-kamera. Hasil eksperimen menunjukkan bahwa kerangka kerja ini mampu menahan rata-rata penggunaan CPU hingga 44.77%, RAM sebanyak 3.16%, dan disk storage sebanyak 6.93% lebih baik dibandingkan dengan kondisi tanpa menggunakannya.
========================================================================================================================
Mining video streams is resource-demanding because it requires substantial resources to explore and extract valuable information from ongoing video data streams. This situation can lead to resource scarcity, causing the system to operate less effectively. One approach to meet this demand is to allocate more resources, but this can increase costs.
This study develops a resource-aware framework for object detection in multi-camera video streaming environments. The Algorithm Granularity Settings (AGS) are used to adapt resource usage across various system components (CPU, RAM, and disk storage) to maintain system stability and performance efficiency. The reliability of this framework is tested for human detection using a deep learning algorithm (YOLOv7-tiny) with input from multi-camera video streams. The experimental results show that this framework is capable of maintaining average CPU usage up to 44.77%, RAM usage by 3.16%, and disk storage by 6.93% better compared to the condition without using it.

Item Type:	Thesis (Masters)
Uncontrolled Keywords:	algorithm granularity settings, data stream mining, deep learning, deteksi objek, object detection, resource-aware.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA402 System analysis. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) Q Science > QA Mathematics > QA76.9.D343 Data mining. Querying (Computer science) T Technology > T Technology (General) > T57.5 Data Processing T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7871.674 Detectors. Sensors
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55101-(S2) Master Thesis
Depositing User:	Yasinta Romadhona
Date Deposited:	06 Aug 2024 05:55
Last Modified:	06 Aug 2024 05:56
URI:	http://repository.its.ac.id/id/eprint/111335

Actions (login required)

View Item