Analisis Sistematis Algoritma Convolutional Neural Network untuk Klasifikasi Gambar Bokeh dan Blur: Tinjauan Literatur

Alif Chandra Wijaya (Universitas Pendidikan Indonesia)
Arditya Baskara Mahbubi (Universitas Pendidikan Indonesia)
Miftah Fauzi Januarta (Universitas Pendidikan Indonesia)
Tasya Syabila (Universitas Pendidikan Indonesia)
Diky Zakaria (Universitas Pendidikan Indonesia)

Jurnal JTIK (Jurnal Teknologi Informasi dan Komunikasi)
Vol 9 No 4 (2025): OCTOBER-DECEMBER 2025

The classification of bokeh and blur images is a challenge in Computer Vision, often addressed using Convolutional Neural Networks (CNNs). This study conducts a Systematic Literature Review (SLR) on 23 articles from Scopus, ScienceDirect, and Google Scholar, with inclusion criteria covering the 2014–2024 publication period, CNN as the primary method, and publication in peer-reviewed journals or conferences (60.87% from scientific journals). The analysis reveals that ResNet and VGG models achieve >90% accuracy, yet still face challenges related to dataset size, computational requirements, and the lack of statistical comparisons across models. This study identifies opportunities for further development through transfer Learning, lightweight models such as MobileNet, and more comprehensive statistical analysis to enhance image classification efficiency across various applications, including digital photography, medical imaging, and security systems.