2-D slices and Metadata from X-ray μCT Images

The original volumetric datasets were published by M. Hlobil and I. Kumpová in Data Brief (Vol. 46, p. 108903, 2023, doi: 10.1016/j.dib.2023.108903). The present dataset was prepared to support temporal deep‑learning research on hydration progression.

Each 3‑D region of interest (ROI) was converted into a series of 2‑D cross‑sectional slices representing different hydration ages, cement fineness levels, and water‑to‑cement ratios. Corresponding binary masks were generated through semantic segmentation to delineate microstructural features relevant to hydration analysis. Metadata files include ROI identifiers, imaging parameters, and temporal pairing information used for Siamese network training.