🌟 Super-Resolution Showcase

A visual benchmark of four deep learning models for Single Image Super-Resolution (SISR) using OpenCV's DNN module. This project compares model quality, runtime, and scale factors in a clear, reproducible format.

🔍 Overview

This project demonstrates the effectiveness of:

EDSR
ESPCN
FSRCNN
LapSRN

Each model is tested at 4× upscaling. Output comparisons help you evaluate both visual clarity and performance runtime.

💡 While all scale models (x2, x3, x4) are included in the models/ folder for reference, the demo scripts use only x4 models for consistency.
🔸 LapSRN does not support x3, but includes a unique x8 model.

🚀 Key Features

🔧 Easy-to-use demos for all 4 models
📊 Runtime comparison for each model
🖼️ Side-by-side visual quality output
🗂️ Modular, reusable utils structure

🧠 How It Works

Load the super-resolution model using OpenCV’s DNN module
Upscale a given low-resolution input
Compare results visually and numerically
Save all outputs in the assets/ folder

📂 Folder Structure

super-resolution-showcase/
│
├── assets/                           # Output, runtime and quality samples, banner
│   ├── quality_comparison.png
│   ├── runtime_chart.png
│   ├── model_output.png
│   └── banner.png
│
├── input/                            # Low-res input images
│   ├── bike-200.png
│   ├── car.jpg
│   ├── car-low.jpg
│   ├── kitten-200.png
│   └── path-low.jpg
│ 
├── output/                           # All program output images are saved here
│
├── ground_truth/                     # High-res originals for quality reference
│   ├── bike-800.png
│   ├── car-high.jpg
│   └── path-high.jpg
│ 
├── models/                           # Pre-trained model files (.pb)
│   ├── edsr/    
│   │   ├── EDSR_x2.pb
│   │   ├── EDSR_x3.pb
│   │   └── EDSR_x4.pb
│   │ 
│   ├── espcn/    
│   │   ├── ESPCN_x2.pb
│   │   ├── ESPCN_x3.pb
│   │   └── ESPCN_x4.pb
│   │ 
│   ├── fsrcnn/  
│   │   ├── FSRCNN_x2.pb
│   │   ├── FSRCNN_x3.pb
│   │   └── FSRCNN_x4.pb
│   │ 
│   └── lapsrn/   
│       ├── LapSRN_x2.pb
│       ├── LapSRN_x4.pb
│       └── LapSRN_x8.pb
│
├── utils/
│   ├── visualizer.py
│   └── superres_utils.py            # Core model loader and inference logic
│
├── demo_all_models.py               # Run all models on sample inputs
├── compare_runtime.py               # Time performance of each model
├── compare_quality.py               # Save visual outputs for comparison
│ 
├── requirements.txt
├── .gitignore
├── LICENSE
└── README.md

🧪 Demo Scripts

Run all models:

python demo_all_models.py

Compare runtime:

python compare_runtime.py

Compare visual outputs:

python compare_quality.py

📁 All outputs are saved inside the output/ folder.

🖼️ Sample Output Gallery

⚡ Super resolution model output

📊 Runtime comparison

🖼️ visual analysis

📦 Requirements

Install required libraries:

pip install -r requirements.txt

🪪 License

This project is open-source under the MIT License

🙋‍♀️ Author

Heleena Robert
GitHub

HeleenaRobert/super-resolution-showcase