This project aims to uncover the palimpsests and obscured texts hidden beneath damaged manuscripts. To achieve this, the aim is to develop methods for palimpsest text separation and handwriting text recognition (HTR) using deep neural networks. Additionally, the project investigates the integration of HTR models with a suitable language model to further enhance the quality of text transcriptions. Our research group is building a multispectral imaging (MSI) system that will be used to capture some real-world images. The experimental set up consists of public benchmark datasets: a synthetic version of the MNIST dataset, where under-text is overlaid with randomly distorted characters from standard English fonts, and the Archimedes dataset, the only known palimpsest character dataset with ground-truth annotations. The ultimate goal is to render the once barely legible remnants of our medieval past accessible once again.