This paper presents a novel technical pipeline for the 3D reconstruction of archaeological artifacts from minimal visual data, specifically single-view image, using Artificial Intelligence-based algorithms. The study addresses a challenging case involving Coptic board games. Due to the inaccessibility of the object caused by different reasons, standard photogrammetric and other survey methods were unfeasible. Leveraging recent advances in computer vision, a deep learning approach derived from the LRM-NeRF family was selected for its superior speed, mesh quality, and generalization performance on out-of-distribution datasets. The reconstruction process included image preprocessing, mesh generation, and geometric optimization using open-source tools. Validation was conducted through metric comparison with similar artifacts held in major museums and expert-in-the-loop reviews. Despite minor texture imperfections due to input limitations, the resulting models proved coherent and analytically reliable. The methodology demonstrates the applicability of 3D reconstruction in low-data heritage scenarios, offering a fully open-source, reproducible and scalable solution for the digital recovery of inaccessible cultural objects. This work not only contributes to the digital preservation of a largely undocumented ludic artifact but also supports the analysis of ancient gameplay dynamics within the broader context of intangible cultural heritage.