Machine Learning and Deep Networks for Additive Wafer Defect Detection: A Concise Study

The mounting demand to maintain wafer quality has made defect identification a critical phase in the chipmaking process. With smaller feature sizes, technology is getting more complicated. Conventional inspection devices or human expertise might become cumbersome due to time spent scanning and altering system settings, associated expenses, and being constrained to a specific known pattern. Machine and Deep Learning-based technologies furnish faster and more accurate results in defect classification as a cost-effective and optimal alternative. However, one must employ a proper algorithm in an industrial or academic setting to avoid these concerns. Recent techniques in defect identification are investigated in this paper, along with their experimental results, utilities, and drawbacks. A well-ordered comparative evaluation of their successful outcomes, constraints, and research barriers will assist practitioners in selecting more appropriate techniques.