Designing novel materials and processes requires a deep understanding of the (micro-)structure-property linkage. Most of the existing computational models require three-dimensional (3D) instances of the materials’ microstructure which are referred to as representative volume elements (RVE). Experimental methods cannot produce these RVEs without significant effort and costs which makes them unfeasible for high-throughput simulations. This rises the demand for microstructure characterization and reconstruction (MCR) methods to generate sets of statistically relevant RVEs with accurate representation of the underlying descriptor. A crucial application of descriptor-based MCR is to extract microstructure descriptors from 2D techniques and optimizes a 3D RVE with respect to the descriptors of interest. MCRpy is a tool that has proven to overcome the limitations of conventional MCR due to the integration of higher-dimensional differentiable descriptors and gradient-based optimization. Seibert et al. have demonstrated the capability of MCRpy to reconstruct grain boundary network of polycrystals [1]. However, to be able to use these reconstructions it is necessary to rigorously connect open grain boundaries to identify grains and mark them using labelling cluster techniques [2]. In this work, we introduce a novel strategy built on MCRpy to generate RVEs from grain boundary networks using artificial neural network as described in Latka et al. [3]. Additionally, we optimize the texture of the RVE by assigning crystallographic orientations to the identified grains. Finally, a case study based on a set of benchmark microstructures is built which exhibit features that pose a challenge to conventional MCR methods. REFERENCES [1] Seibert, Raßloff, Ambati, Kästner, Descriptor-based reconstruction of three-dimensional microstructures through gradient-based optimization, Acta Materialia, 2022. [2] Hoshen, Kopelmann, Percolation and cluster distribution. I. Cluster multiple labeling technique and critical concentration algorithm, Physical Review B, 1976. [3] Latka, Doškář, Zeman, Microstructure reconstruction via artificial neural networks: a combination of causal and non-causal approach, Acta Polytechnica CTU Proceedings, 2022.