Revive Vintage Photos with AI: Guide to Bringing-Old-Photos-Back-to-Life

Today we recommend the project “Bringing-Old-Photos-Back-to-Life”, an AI‑driven tool that restores and revitalizes old photographs.

The project provides a Colab demo where users can upload their own images and see the restoration results.

Colab Demo URL: https://colab.research.google.com/drive/1NEm6AsybIiC5TwTU_4DqDkQO0nFRB-uA?usp=sharing

Requirements

The code has been tested on Ubuntu with Nvidia GPUs and CUDA installed.

Python version 3.6 or higher is required.

How to Use

Full Pipeline

Run the following command to install the pretrained model and restore images without scratches:

python run.py --input_folder [test_image_folder_path] \
               --output_folder [output_path] \
               --GPU 0

For images with scratches, add the --with_scratch flag:

python run.py --input_folder [test_image_folder_path] \
               --output_folder [output_path] \
               --GPU 0 \
               --with_scratch

Scratch Detection

The scratch‑detection dataset is not released, but the pretrained model can be used to generate labels for collected images:

cd Global/
python detection.py --test_path [test_image_folder_path] \
                    --output_dir [output_path] \
                    --input_size [resize_256|full_size|scale_256]

Global Restoration

The project introduces a triple‑domain conversion network to address both structural and non‑structural degradation in old photos.

Face Enhancement

A progressive generator refines facial regions of old photos; further details are available in the repository’s /Face_Enhancement folder.

Open‑source repository: https://github.com/microsoft/Bringing-Old-Photos-Back-to-Life (Microsoft).