ScholarWorks@경상국립대학교

초록

Deep learning approach has revolutionized image processing, offering capabilities that go far beyond traditional methods. This paper reviews how deep learning methods (from early breakthroughs to today’s cutting-edge models) have evolved to better handle complex visual data, improving efficiency, generalization, and robustness across many applications. Finally, we suggest future directions such as combining deep learning with quantum or neuromorphic computing, using federated learning for data privacy, and integrating edge computing and explainable artificial intelligence to tackle scalability and interpretability challenges. The survey highlights the strategic value of the deep learning-based image processing for ultra-fine defect inspection and process control in semiconductor and display manufacturing lines.

키워드