<p>In this work, two different variants of image segmentation are compared to evaluate the use of generalized machine learning models against the accuracy of bespoke models to further their use for the analysis of microstructure images with multiple phases. The results from the analysis are then used to evaluate the effect of different iron contents and the presence or absence of convection on the formation of the microstructure with emphasis on the development of the intermetallic phases in technical aluminum alloys. To this end, the study focuses on aluminum-silicon base cast alloys with high iron content directionally solidified under microgravity conditions, with additional controlled convection created by a rotating magnetic field. Optical microscopy images from the different processing zones are then used to train the different chosen models, which are afterwards used to segment and analyze the microstructures. Key results include the evaluation of the effects of convection and iron content on several parameters describing the different intermetallic phases as well as the comparison of the models.</p>