晶圆缺陷数据集

本研究展示了一套从DRAM生产线后端工艺线阶段采集的缺陷复查扫描电镜图像数据集，使用SEM机台拍摄。数据集包含1150张SEM图像，涵盖100多种不同缺陷类别。这些于2023至2024年间采集的图像经专家精心分类，确保缺陷归类的准确性与可靠性。该数据集作为缺陷分析的宝贵资源，对推动半导体产业研发具有重大潜力。通过公开该数据集，促进半导体缺陷检测与分析专用算法的开发，推动该专业领域与更广泛图像处理研究的融合

This study presents a dataset of defect-review scanning electron microscopy (SEM) images collected from the backend process stage of DRAM production lines, captured using dedicated SEM inspection equipment. The dataset comprises 1,150 SEM images covering over 100 distinct defect categories. All images acquired between 2023 and 2024 have been meticulously classified by domain experts to guarantee the accuracy and reliability of defect categorization. As a valuable resource for semiconductor defect analysis, this dataset possesses considerable potential to advance research and development within the semiconductor industry. By publicly releasing this dataset, it supports the development of specialized algorithms for semiconductor defect detection and analysis, and fosters the integration of this specialized field with broader image processing research.

搜集汇总

背景与挑战

背景概述

晶圆缺陷数据集是一套从DRAM生产线后端工艺线阶段采集的缺陷复查扫描电镜图像数据集，包含1150张SEM图像，涵盖100多种不同缺陷类别。该数据集由长鑫存储技术有限公司发布，适用于半导体产业研发和图像处理研究。