Analisis Perkembangan Deep Learning Pada Pengolahan Citra Digital

Abstract

The rapid advancement of deep learning has significantly transformed the field of digital image processing. Various architectures, including Convolutional Neural Networks (CNN), ResNet, EfficientNet, and Vision Transformers (ViT), have been developed to improve accuracy and computational efficiency in image classification, object detection, image segmentation, and pattern recognition tasks. This study aims to analyze the development of deep learning in digital image processing using a Systematic Literature Review (SLR) approach. The research methodology follows the PRISMA 2020 guidelines, consisting of identification, screening, eligibility assessment, and synthesis of scientific articles retrieved from Scopus, IEEE Xplore, ScienceDirect, SpringerLink, ACM Digital Library, and Google Scholar. A total of 50 articles that met the inclusion criteria were analyzed descriptively to identify publication trends, commonly used architectures, application domains, strengths, limitations, and future research opportunities. The findings indicate that CNN remains the most widely adopted architecture, while Vision Transformers and hybrid CNN–Transformer models have gained increasing attention in recent years due to their superior performance in various computer vision tasks. The healthcare sector represents the most prominent application area, followed by transportation, agriculture, security, and remote sensing. Despite its outstanding performance, deep learning still faces challenges related to large-scale dataset requirements, high computational costs, and limited model interpretability. This literature review provides a comprehensive overview of the current state of deep learning in digital image processing and offers valuable insights for future research and practical applications.