🎓 Intel Unnati 2025 - Image Super-Resolution Project

Knowledge Distillation for Image Super-Resolution

📊 Project Overview

This project implements a lightweight image super-resolution model using knowledge distillation techniques as part of the Intel Unnati Industrial Training Program 2025.

🎯 Key Achievements

• ✅ 92.7% SSIM Score (Target: >90%)
• ✅ 483K Parameters (Lightweight model)
• ✅ Real-time Processing up to 1920×1080
• ✅ Knowledge Distillation from SwinIR teacher model

Dataset Used is uploaded in google Drive Link : https://drive.google.com/drive/folders/1S0-4pJsoOoKeGSiSCTVQCso6eTyZbri9?usp=sharing

🏗️ Project Structure

Intel_project_final_version/
├── Intel_Unnati_Project.ipynb    # Main implementation notebook
├── DIV2K_train_HR/                # Dataset (800 high-res images)
├── student_models/                # Trained model weights
├── TeacherModel/                  # SwinIR teacher model
├── data/                          # Processed training data
├── 100_enhanced_results.png       # Visual results
└── README.md                      # This file

🔧 Technical Specifications

• Dataset: 800 DIV2K high-resolution images
• Architecture: Enhanced CNN with 6 residual blocks
• Training: Knowledge distillation with L1+MSE loss
• Performance: 92.7% average SSIM, 28.5 dB PSNR
• Scalability: 100×100 to 1920×1080 resolution

📋 Requirements

torch>=1.9.0
torchvision>=0.10.0
opencv-python>=4.5.0
scikit-image>=0.18.0
matplotlib>=3.3.0
numpy>=1.21.0
Pillow>=8.3.0

🚀 Usage

1. Open Intel_Unnati_Project.ipynb in Jupyter/VS Code
2. Run cells sequentially to:
   • Set up environment
   • Load and preprocess data
   • Train the model
   • Evaluate performance
   • Generate visual results

📈 Results

• SSIM Score: 92.7% (exceeds 90% target)

• Model Size: 1.85 MB

• PSNR Score: 28.5 dB (exceeds 25 dB target)

• Resolution: 1920×1080 (real-time processing)

• Model Parameters: 483K

• Knowledge Distillation: Implemented

• Teacher Model: SwinIR (Swin Transformer)

• Success Rate: 87% of images achieve >90% SSIM

🏆 Intel Unnati Requirements Met

• ✅ Knowledge distillation implementation
• ✅ 100×100 patch training
• ✅ 1920×1080 scalability testing
• ✅ Performance target exceeded
• ✅ Lightweight model achieved

👨‍💼 Author

Team: Code triplets
Program: Intel Unnati Industrial Training 2025
Project: Knowledge Distillation for Image Super-Resolution

---

Project Status: ✅ SUCCESSFULLY COMPLETED