Comparison of F1-scores from human assignments or automated focus assignment to other experimenters^a.

aThe average F1 score between experimenters across 21 images using the original manually assigned foci from each experimenter (‘Original’) or the automated focus assignment trained across a dataset (‘Algorithm’) when compared against the ‘ground truth’, defined by another experimenter. Matches were iteratively assigned using the closest pairs within 8 pixels. Values are shown to three significant figures. Comparison of F1-scores from human assignments or automated focus assignment to other experimentersa.本研究针对21幅图像，对比了两种方式下不同实验者间的平均F1分数：一种是使用每位实验者原始手动标注的焦点（记为"Original"），另一种是使用在数据集上训练得到的自动焦点分配方法（记为"Algorithm"），对比基准为另一位实验者定义的真实标签（ground truth）。匹配过程通过选取8像素范围内的最邻近匹配对进行迭代分配，结果保留三位有效数字。本对比展示了人类手动标注与自动焦点分配的F1分数，与其他实验者标注结果的对比情况。