ChronoVision is a deep learning pipeline designed to restore and colorize historical black-and-white photographs by converting low-quality grayscale inputs into high-resolution color outputs. By integrating a custom fine-tuned U-Net (EfficientNet-B2) for chrominance prediction with specialized restoration models (GFPGAN and Real-ESRGAN), the system mitigates semantic ambiguity and hallucinates missing textural details. This project aims to make history more accessible for archivists and educators by bridging the gap between deterministic colorization and generative facial restoration.
Fullerite/chrono-vision
Folders and files
| Name | Name | Last commit date | ||
|---|---|---|---|---|