🧠 MachineLearning

📚 Educational Machine Learning Repository - A comprehensive collection of deep learning and machine learning implementations designed for educational purposes. This repository features various neural network architectures and ML projects built primarily in PyTorch, with a focus on demonstrating fundamental concepts, training methodologies, and architectural principles of modern AI systems. Each project is implemented from scratch to provide clear, step-by-step examples that facilitate understanding of how neural networks and computer vision models operate at their core, prioritizing educational value and code clarity over production optimization.

🤖 Models

🖼️ Deblur

Image Deblurring Network - A CNN-based deep learning model for removing Gaussian blur from images.

• Architecture: 7-layer CNN with residual connections (1.33M parameters)
• Training Approach: Single-image overfitting for demonstration purposes
• Loss Functions: Combined MSE + VGG19 Perceptual Loss
• Features: Mixed precision training, gradient clipping, TF32 acceleration
• Input/Output: 512×512 RGB images
• Use Case: Educational demonstration of image restoration techniques

The model learns to map blurred images back to sharp versions using a symmetric encoder-decoder architecture with skip connections to preserve original image information.

🔍 ImageRecognition

Multi-Dataset Image Recognition Network - A comprehensive CNN for classifying hundreds of classes across multiple datasets.

• Architecture: Wide CNN with 4 blocks (128→256→512→1024 filters, ~83M parameters)
• Training Approach: Multi-dataset unified training across 10+ popular CV datasets
• Datasets: MNIST, Fashion-MNIST, CIFAR-10/100, SVHN, STL-10, Food-101, Flowers-102, Oxford Pet, Caltech-101/256
• Features: Mixed precision training, extensive data augmentation, automatic checkpointing
• Input/Output: 64×64 RGB images → hundreds of class predictions with confidence scores
• Use Case: Comprehensive multi-domain image classification system

The model combines multiple datasets into a unified recognition system capable of identifying digits, fashion items, objects, food, flowers, pets, and more through a single neural network.

🚀 Setup

Quick Start

This project uses the Scripts/ directory for easy environment setup across different platforms. The setup process will create a virtual environment and install all required dependencies automatically.

Windows Users:

# Run from project root directory
Scripts\WinSetup.bat

Linux/macOS Users:

# Make executable and run from project root
chmod +x Scripts/LinuxSetup.sh
./Scripts/LinuxSetup.sh

Daily Usage (Activate Virtual Environment):

Windows (Command Prompt):

Scripts\EnvActivate.bat

Windows (PowerShell):

.\Scripts\EnvActivate.ps1

Linux/macOS:

chmod +x Scripts/EnvActivate.sh  # First time only
./Scripts/EnvActivate.sh

What the Setup Does:

• 🐍 Auto-detects Python: Works with python, py, or python3 commands
• 📦 Creates Virtual Environment: Isolated .venv folder in project root
• ⬇️ Installs Dependencies: All packages from requirements.txt
• ✅ Verification: Confirms successful setup with clear messaging

For detailed setup instructions, troubleshooting, and advanced usage, see Scripts/README.md.

📋 Requirements

The following Python packages are automatically installed during setup:

• torch>=2.0.0 - PyTorch deep learning framework
• torchvision>=0.15.0 - Computer vision utilities and models
• Pillow>=9.0.0 - Python Imaging Library for image processing
• numpy>=1.21.0 - Numerical computing and array operations
• matplotlib>=3.5.0 - Plotting and data visualization
• pandas>=1.3.0 - Data manipulation and analysis
• scipy>=1.7.0 - Scientific computing and optimization
• argparse - Command-line argument parsing

System Requirements:

• Python 3.8+ (Python 3.9+ recommended)
• CUDA-compatible GPU (optional, but recommended for faster training)
• 4GB+ RAM (8GB+ recommended for GPU training)

💻 Hardware Recommendations

• GPU: CUDA-compatible GPU with 4GB+ VRAM for optimal performance
• CPU: Fallback support available but significantly slower
• Memory: 8GB+ system RAM recommended
• Storage: SSD recommended for faster data loading

📁 Project Structure

MachineLearning/
├── .venv/                    # Virtual environment (created during setup)
├── Scripts/                  # Setup and activation scripts
│   ├── WinSetup.bat         # Windows environment setup
│   ├── LinuxSetup.sh        # Linux/macOS environment setup
│   ├── EnvActivate.*        # Environment activation scripts
│   └── README.md            # Detailed setup documentation
├── Deblur/                  # Image deblurring project
│   ├── Deblur.py           # Main training script
│   ├── PerceptualLoss.py   # VGG19 perceptual loss implementation
│   ├── data/               # Training images directory
│   └── README.md           # Project-specific documentation
├── requirements.txt         # Python dependencies
├── LICENSE                 # Project license
└── README.md               # This file

🎯 Getting Started

1. Clone the repository
2. Run the appropriate setup script from the Scripts/ directory
3. Activate the virtual environment using the activation scripts
4. Navigate to a project folder (e.g., Deblur/)
5. Follow project-specific README for detailed instructions

🔮 Future Projects

This repository will continue to grow with additional machine learning projects covering:

• Generative models (GANs, VAEs)
• Natural Language Processing
• Reinforcement Learning
• Object Detection and Segmentation
• Transfer Learning examples

⚠️ Educational Notice

These projects are designed for learning purposes and may not be optimized for production use. They emphasize clarity and educational value over performance optimization.

📄 License

This project is for educational purposes. Please refer to the LICENSE file for usage terms.