Preserving Oriya Manuscript Integrity: Transformer-Based Self-Supervised Imputation

Historical manuscript conservation and restoration are essential for preserving cultural heritage. Manuscripts offer proof of ideas, facts, and theories that are important for human growth, directing the next generation and need to be kept for the benefit of future generations. This study present a novel method to handle missing information in Oriya manuscripts images: Transformer-based Self-Supervised Imputation. The proposed approach predicts and restores missing textual and visual components of manuscripts by utilizing a Transformer model in conjunction with a hybridization of advanced imputation approaches. In order to demonstrate TSSI's effectiveness in maintaining manuscript integrity, the novel method had been compared to conventional imputation techniques like Mean, Median, Mode, KNN, and MICE by calculating quantitative measures such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). This paper also present the potential of TSSI to precisely reconstruct lost material through visualization and comparison, hence aiding in the preservation of Oriya cultural artifacts.