arthistoricum.net

In the area of conserving cultural heritage and enhancing its accessibility for individuals with disabilities, we introduce a methodology designed to extract intricate details from portrayed coins and to produce three-dimensional (3D) representations of the artifacts depicted on them. Our approach consists of three distinct phases, each executed asynchronously and offline. The initial phase is dedicated to the mitigation of image noise in the coin's visual representation, stemming from various factors such as suboptimal photography techniques or imperfections during the manufacturing process. The subsequent phase involves the derivation of segmentation masks and an evaluation of the attained level of detail pertinent to the specific issue at hand. The concluding phase utilizes the extracted objects to formulate 3D models. Throughout these stages, we employ cutting-edge deep learning models that have undergone meticulous fine-tuning.