ReviveAI ✨

ReviveAI Logo

Restore your memories. AI-powered image deblurring, sharpening, and scratch removal.

📖 About ReviveAI

ReviveAI leverages the power of Artificial Intelligence to breathe new life into your old or degraded photographs. Whether it's blurriness from camera shake, general lack of sharpness, or physical damage like scratches, ReviveAI aims to restore clarity and detail, preserving your precious moments.

This project utilizes state-of-the-art deep learning models trained specifically for image restoration tasks. Our goal is to provide an accessible tool for enhancing image quality significantly.

🔥 Key Features

Features

✅ Completed - Image Sharpening: Enhances fine details and edges for a crisper look.
✅ Completed - Scratch Removal: Intelligently detects and inpaints scratches and minor damages on photographs.
🛠️ Work-in-progress - Image Colorization(Coming Soon): Adds realistic color to grayscale images.

✨ Before & After Showcase

See the results of ReviveAI in action!

Examples	Task Performed
	Image Sharpening
	Image Sharpening
	Scratch Removal
	Scratch Removal

🛠️ Tech Stack

📊 Implementation Status

Track the development progress of ReviveAI's key features and components:

Feature / Component	Status	Notes / Remarks (Optional)
Image Deblurring/Sharpening	✅ Completed	Core model functional
Scratch Removal	✅ Completed	Core model functional
Image Colorization	🚧 In Progress	Model integration underway

🚀 Getting Started

Follow these steps to get ReviveAI running on your local machine or in a Jupyter/Kaggle notebook.

1. Prerequisites

Ensure you have the following installed:

Python 3.8 or above
pip (Python package manager)
Git (for cloning the repository)
Hugging Face CLI (optional)
Jupyter Notebook or run on Kaggle / Google Colab

2. Clone the Repository

git clone https://github.com/Zummya/ReviveAI.git
cd ReviveAI

3. Set Up the Environment

We recommend using a virtual environment:

python -m venv env
source env/bin/activate  # On Windows: env\Scripts\activate
pip install -r requirements.txt

🎯 Load Pretrained Models

All models are hosted on the Hugging Face Hub for convenience and version control.

🔹 Load Image Sharpening Model

from huggingface_hub import hf_hub_download
from tensorflow.keras.models import load_model

model_path = hf_hub_download(
    repo_id="Sami-on-hugging-face/RevAI_Deblur_Model", 
    filename="SharpeningModel_512_30Epochs.keras"
)
model = load_model(model_path, compile=False)

🔹 Load Scratch Removal Model

from huggingface_hub import hf_hub_download
from tensorflow.keras.models import load_model

model_path = hf_hub_download(
    repo_id="Sami-on-hugging-face/RevAI_Scratch_Removal_Model", 
    filename="scratch_removal_test2.h5"
)
model = load_model(model_path, compile=False)

📁 Folder Structure

ReviveAI/
│
├── README.md
├── .gitignore
├── requirements.txt
│
├── models/
│   ├── sharpening_model.txt         # Hugging Face URL
│   └── scratch_removal_model.txt    # Hugging Face URL
│
├── notebooks/
│   ├── scratch_removal_notebook.ipynb
│   └── sharpening_model_notebook.ipynb
│
├── before_after_examples/
│   ├── sharpening/
│   └── scratch_removal/
│
├── assets/
│   └── revive banner.png, showcase images etc.

🧪 Training & Running the Models

ReviveAI includes end-to-end Jupyter notebooks that allow you to both train the models from scratch and test them on custom images.

📘 Available Notebooks

Notebook	Description
`sharpening_model_notebook.ipynb`	Train the sharpening (deblurring) model + Run predictions
`scratch_removal_notebook.ipynb`	Train the scratch removal model + Run predictions

💡 Notebook Features

Each notebook includes:

🧠 Model Architecture
🔁 Data Loading & Preprocessing
🏋️ Training Pipeline (with adjustable hyperparameters)
💾 Saving & Exporting Weights
🔍 Evaluation
🖼️ Visual Demo on Custom Images

🖼️ Quick Test Function (for inference)

To run a prediction on a new image (after training or loading a model), use:

def display_prediction(image_path, model):
    import cv2
    import matplotlib.pyplot as plt
    import numpy as np

    img = cv2.imread(image_path)
    img = cv2.resize(img, (256, 256)) / 255.0
    input_img = np.expand_dims(img, axis=0)
    predicted = model.predict(input_img)[0]

    plt.figure(figsize=(10, 5))
    plt.subplot(1, 2, 1)
    plt.imshow(img[..., ::-1])
    plt.title("Original Input")
    plt.axis("off")

    plt.subplot(1, 2, 2)
    plt.imshow(predicted)
    plt.title("Model Output")
    plt.axis("off")

    plt.show()

Run the function like this:

display_prediction("your_image_path.jpg", model)

✅ Tip: If you don't want to train from scratch, you can directly load pretrained weights from Hugging Face (see 🎯 Load Pretrained Models) and skip to the testing section.

istevit
/

ReviveAI