Two types of convolutional neural network (CNN) models, a discrete classification network and a continuous regression network, were trained to determine local sample thickness from convergent beam diffraction (CBED) patterns of SrTiO3 collected in a scanning transmission electron microscope (STEM) at atomic column resolution. Acquisition of atomic resolution CBED patterns for this purpose requires careful balancing of CBED feature size in pixels, acquisition speed, and detector dynamic range. The training datasets were derived from multislice simulations, which must be convolved with incoherent source broadening. Sample thicknesses were also determined using quantitative high-angle annular dark-field (HAADF) STEM images acquired simultaneously. The regression CNN performed well on sample thinner than 35 nm, with 70% of the CNN results within 1 nm of HAADF thickness, and 1.0 nm overall root mean square error between the two measurements. The classification CNN was trained for a thicknesses up to 100 nm and yielded 66% of CNN results within one classification increment of 2 nm of HAADF thickness. Our approach depends on methods from computer vision including transfer learning and image augmentation.

中文翻译：

---

使用卷积神经网络的原子分辨率会聚束电子衍射分析

训练了两种类型的卷积神经网络 (CNN) 模型，即离散分类网络和连续回归网络，以根据在原子级扫描透射电子显微镜 (STEM) 中收集的 SrTiO3 的会聚光束衍射 (CBED) 图案确定局部样品厚度列分辨率。为此目的采集原子分辨率 CBED 模式需要仔细平衡 CBED 特征尺寸（以像素为单位）、采集速度和检测器动态范围。训练数据集来自多层模拟，它必须与不相干的源扩展进行卷积。还使用同时获取的定量高角度环形暗场 (HAADF) STEM 图像确定样品厚度。回归 CNN 在小于 35 nm 的样品上表现良好，70% 的 CNN 结果在 HAADF 厚度的 1 nm 以内，两次测量之间的总体均方根误差为 1.0 nm。分类 CNN 针对高达 100 nm 的厚度进行了训练，并在 2 nm 的 HAADF 厚度的一个分类增量内产生了 66% 的 CNN 结果。我们的方法依赖于计算机视觉的方法，包括迁移学习和图像增强。