2025_SunHao_SEM_depth_estimation-TIM

This dataset complements our study on monocular depth estimation of scanning electron microscope (SEM) images driven by physics-informed priors, which was presented in the associated manuscript. It comprises two complementary subsets: (1) a synthetic SEM image\u2010depth pair ensemble generated via our physically-based simulator, covering a broad domain of material types, imaging conditions and depth ranges; and (2) a real-world validation set of SEM micrographs captured under controlled conditions alongside ground-truth depth measurements, used to benchmark our deep learning model\u2019s performance. The dataset enables reproducibility of the training process, supports evaluation of generalization across imaging scenarios, and facilitates analysis of the effect of physics prior integration on depth-map quality. The files are accompanied by code for pre-processing, model inference and statistical performance evaluation. We anticipate this dataset to be a useful resource for the semiconductor imaging, computer vision and microscopy-metrology communities seeking to advance single-image depth estimation techniques under high-magnification, low-contrast conditions.