Implementasi SlowFast Network dalam Klasifikasi Kejadian Hampir Kecelakaan

Teja, Andika Rahman (2025) Implementasi SlowFast Network dalam Klasifikasi Kejadian Hampir Kecelakaan. Project Report. [s.n.], [s.l.]. (Unpublished)

Text ​5025221022-Project_Report.pdf - Accepted Version Download (2MB)

Abstract

Teknologi mobil swakemudi terus berkembang dengan integrasi kecerdasan buatan untuk meningkatkan keselamatan berkendara. Salah satu pendekatan yang digunakan adalah penerapan arsitektur SlowFast Network guna memahami informasi spasial dan temporal dari rekaman video untuk mendeteksi kejadian akan kecelakaan. Pada penelitian ini mengevaluasi performa lima konfigurasi model SlowFast Network pada data video kecelakaan DADA-2000. Hasil pengujian menunjukkan bahwa konfigurasi SlowFast 4x16 mencapai akurasi tertinggi dengan Top-1 Accuracy sebesar 68.24%. Temuan ini menegaskan efektivitas SlowFast Network dalam mengolah informasi secara simultan dan dapat mendukung pengembangan sistem deteksi kecelakaan yang lebih akurat dan responsif. Untuk dapat meningkatkan potensi model SlowFast Network sebagai sistem deteksi kecelakaan, penulis menyarankan untuk melakukan hyperparameter tuning terhadap tiap konfigurasi model SlowFast Network dan menerapkan teknik cross-validation untuk evaluasi model yang lebih andal.
=====================================================================================================================================
Self-driving-car technology is advancing rapidly, integrating artificial intelligence to boost road safety. One promising approach is the SlowFast Network architecture, which jointly captures spatial and temporal information from video footage to anticipate potential accidents. This study evaluates five SlowFast Network configurations on the DADA-2000 traffic-accident video dataset. The experiments reveal that the SlowFast 4×16 configuration achieves the best performance, attaining a Top-1 accuracy of 68.24 percent. These results underscore the network’s effectiveness in processing spatial–temporal cues simultaneously and its potential for enabling more accurate and responsive accident-detection systems. To further enhance the model’s capabilities, we recommend hyperparameter tuning for each configuration and the use of cross-validation techniques for more robust evaluation.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Mobil Swakemudi, SlowFast Network, Deteksi Kecelakaan Self-Driving Car, Accident Detection
Subjects:	Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) > T174.5 Technology--Risk assessment.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Andika Rahman Teja
Date Deposited:	04 Jul 2025 08:08
Last Modified:	04 Jul 2025 08:08
URI:	http://repository.its.ac.id/id/eprint/119369

Actions (login required)

View Item