document-restoration

Change text in scanned documents & photos using the document's own letters — typewriter, printed fonts, handwriting. Python + OpenCV, no GPU.

Line-level restoration of degraded Arabic manuscripts: a U-Net that makes damaged handwriting readable, scored by whether an OCR model can read it.

An automated computer vision pipeline for historical document restoration, featuring classical CV techniques, a trained U-Net deep learning model, and an interactive Streamlit comparison dashboard.

A hybrid restoration framework for ancient documents that combines Deep Learning (U-Net) for structural mask detection with classical geometric transformations (TPS, Hough Transform) to rectify distortions while ensuring 100% historical integrity and verifiability.