Generative AI for digital color restoration in archaeology: Reconstructing lost pigments of the Harappan Civilization

Color plays a crucial role in understanding the materiality, symbolism, and aesthetic identity of ancient artifacts. However, centuries of degradation have caused irreversible pigment loss in archaeological collections, limiting both scholarly interpretation and public appreciation. This study introduces a novel framework that employs Generative Artificial Intelligence (AI) to digitally reconstruct lost pigments in Harappan artifacts, one of the earliest urban cultures of South Asia. Using high-resolution images of terracotta pottery, seals, and figurines from museum and open-access archives, a diffusion-based generative Model was fine-tuned to predict original chromatic values by referencing extant pigment datasets and contextual archaeological metadata. The reconstructed images were then processed through a colorimetric calibration pipeline to ensure perceptual accuracy and consistency across devices. Quantitative evaluations compared generated hues with known pigment spectra, while expert validation from archaeologists assessed visual plausibility. The results demonstrate that generative AI can effectively simulate the original appearance of faded or eroded artifacts, offering new possibilities for virtual restoration, digital storytelling, and educational visualization. Beyond color recovery, the study highlights how AI-driven chromatic reconstruction contributes to heritage preservation and interpretation, providing a reproducible methodology for reimagining ancient aesthetics through computational means. This approach expands the frontier of digital color conservation by bridging heritage science, machine learning, and archaeological visualization within a sustainable, culturally sensitive framework.