Image Deblurring using Generative Adversarial Networks (GAN) 🖼️✨

This project implements an Image Deblurring system using Generative Adversarial Networks (GANs).
The goal is to reconstruct sharp images from motion-blurred inputs by training a generator–discriminator architecture using deep learning.

The implementation is done in Python using TensorFlow/Keras and OpenCV.

---

📁 Project Structure

.
├── image_deblurring_gan.py      # Main training and inference script
├── generated_images/            # Generated images saved during training
├── README.md                    # Project documentation

📌 Project Overview

The project consists of:

• Loading motion-blurred and sharp image pairs

• Preprocessing images (resizing and normalization)

• Training a GAN model

  -- Generator learns to produce sharp images

  -- Discriminator distinguishes real vs generated images

• Generating and saving deblurred outputs

• Testing the trained generator on new blurry images

🧰 Software Requirements

Programming Language

• Python 3.7+

Required Libraries

pip install numpy opencv-python matplotlib tensorflow keras

📊 Dataset

The dataset used for training is sourced from Kaggle:

🔗 Blur Dataset https://www.kaggle.com/datasets/kwentar/blur-dataset?resource=download

🧠 Model Architecture

Generator

• Dense + Reshape layers
• Upsampling using UpSampling2D
• Convolutional layers with Batch Normalization
• Outputs a 64×64×3 RGB image

Discriminator

• Convolutional layers with strided downsampling
• Fully connected classifier
• Binary output (real vs fake)

📈 Training Details

• Image size: 64 × 64
• Latent vector size: 100
• Batch size: 32
• Epochs: 1000
• Loss function: Binary Cross-Entropy
• Optimizer: Adam

⚠️ Notes & Limitations

• This is a basic GAN, not a conditional GAN (cGAN)
• Training stability may vary
• Performance depends heavily on dataset size and quality
• Not optimized for high-resolution images

📜 License

• This project is intended for educational and academic purposes.
• You are free to modify and extend the code.

👤 Author

• Shakthi Bala

---

✅ Why this README is solid

• Matches your actual code
• Clear dataset usage
• Explains GAN logic at the right level
• Recruiter- and academic-friendly
• GitHub-rendering safe

If you want next, I can:

• Convert this to cGAN-based deblurring
• Improve the training loop
• Add PSNR / SSIM evaluation
• Make all your READMEs follow a single personal template

Just say 👍