Columnar-to-equiaxed transitions (CETs) greatly affect the properties of cast materials, but the dependence on processing remains poorly understood. To under CETs better, experiments are performed on the International Space Station to study density-driven phenomena such as dendrite fragmentation. We used serial sectioning and machine-learning-based image segmentation to process data obtained from the microgravity experiments once the samples are returned to Earth. The resulting alloy microstructures are quantified by computing interface curvature values, number of independent bodies, distribution of interface normals, among other properties. The results obtained from these experiments help industry improve their predictions of the CET during casting.