AI-Powered Image Restoration and Enhancement Tool

[ramana2074]

---

name: DL-Simplified Issue Template
about: This is the issue template that is going to be used while creating any issues
in this repo.
title: ''
labels: ''
assignees: ''

---

Deep Learning Simplified Repository (Proposing new issue)

🔴 Project Title :
AI-Powered Image Restoration and Enhancement Tool
🔴 Aim :
This project aims to develop an AI-Powered Image Restoration Tool that revitalizes low-quality historical photos by upscaling, denoising, and adding realistic color.
🔴 Dataset :
Pretrained model
🔴 Approach : Using advanced deep learning techniques, it transforms degraded images into vibrant, high-quality visuals while preserving their historical context. This tool is perfect for heritage conservation, family archiving, and historical documentation.
SRCNN (Super-Resolution Convolutional Neural Network): For image upscaling
DnCNN: For image denoising
Pre-trained colorization models (e.g., U-Net): For adding color to grayscale images

---

📍 Follow the Guidelines to Contribute in the Project :

• You need to create a separate folder named as the Project Title.
• Inside that folder, there will be four main components.
  • Images - To store the required images.
  • Dataset - To store the dataset or, information/source about the dataset.
  • Model - To store the machine learning model you've created using the dataset.
  • requirements.txt - This file will contain the required packages/libraries to run the project in other machines.
• Inside the Model folder, the README.md file must be filled up properly, with proper visualizations and conclusions.

---

🔴🟡 Points to Note :

• The issues will be assigned on a first come first serve basis, 1 Issue == 1 PR.
• "Issue Title" and "PR Title should be the same. Include issue number along with it.
• Follow Contributing Guidelines & Code of Conduct before start Contributing.

---

✅ To be Mentioned while taking the issue :

• Full name : Venkata Ramana Billana
• GitHub Profile Link : https://github.com/ramana2074
• Email ID : billanavenkataramana@gmail.com
• Participant ID (if applicable):
• Approach for this Project : Using advanced deep learning techniques, it transforms degraded images into vibrant, high-quality visuals while preserving their historical context. This tool is perfect for heritage conservation, family archiving, and historical documentation.
• What is your participant role? (Mention the Open Source program) - SWOC

---

Happy Contributing 🚀

All the best. Enjoy your open source journey ahead. 😎

[github-actions]

Thank you for creating this issue! We'll look into it as soon as possible. Your contributions are highly appreciated! 😊

[abhisheks008]

Hi @ramana2074 issue assigned to you.

[Sonmale25]

Full name:

Pranav Prakash Sonmale

GitHub Profile Link:

https://github.com/Sonmale25

Email ID:

sonmalepranav@gmail.com

What is your participant role?

SSOC ’25 Contributor

### Proposed Approach
I will build a pipeline that restores, denoises, upscales, and colorizes historical images using state‑of‑the‑art convolutional architectures:

### Model Components & Order

Denoising: DnCNN to remove sensor noise and scratches.

Super‑resolution (Upscaling): SRCNN to increase resolution by 2× or 4×.

Colorization: U‑Net‑based pre‑trained colorization network to add realistic color while preserving historical detail.

Tech Stack

Framework: Python 3.9+ with PyTorch for model development.

Dependencies:

torch, torchvision

numpy, opencv-python for image I/O & preprocessing

matplotlib or seaborn for visualizations

Pillow for format conversions

Dataset & Pretrained Weights

DnCNN: use official pretrained weights trained on BSD400.

SRCNN: use publicly available weights trained on DIV2K dataset.

Colorization U‑Net: leverage “Richard Zhang et al.” pretrained model from [GitHub/colorful‑image‑colorization].

Custom Fine‑Tuning (Optional): allow fine‑tuning on user‑provided small historical‑photo datasets (e.g. Yorick’s “Historical Photo Restorations” set) for domain adaptation.

Inference Pipeline

Load an input image → apply DnCNN → feed denoised image to SRCNN → pass upscaled output to U‑Net colorizer → save final PNG/JPEG.

Provide a Jupyter notebook (Model/inference_demo.ipynb) demonstrating end‑to‑end results on sample images.

Evaluation & Visualization

Compute PSNR and SSIM against available ground‑truth restorations (if any).

In Model/README.md, include side‑by‑side before/after comparisons, quantitative metrics, and brief conclusions on restoration quality.

Looking forward to contributing—please assign this