Pseudo-Color Mapping for Thermal Images

An academic project implementing pseudo-color mapping on grayscale thermal images, feature extraction, and machine learning classification of temperature categories.

Project Structure

pseudo_color_thermal/
├── data/
│   ├── raw/            # Original FLIR grayscale thermal images
│   ├── processed/      # Preprocessed images
│   └── samples/        # Sample images for quick testing
├── src/
│   ├── preprocess.py   # Image loading & preprocessing
│   ├── colormap.py     # Pseudo-color mapping
│   ├── features.py     # Feature extraction
│   ├── model.py        # ML model training & evaluation
│   ├── visualize.py    # Visualization utilities
│   └── utils.py        # Helper functions
├── outputs/
│   ├── images/         # Saved processed images
│   ├── plots/          # Saved plots & charts
│   └── metrics/        # Saved evaluation metrics (CSV/JSON)
├── models/             # Saved trained models
├── notebooks/          # Jupyter notebooks (optional)
├── logs/               # Runtime logs
├── main.py             # Entry point — runs the full pipeline
├── config.py           # All configurable parameters
└── requirements.txt

Setup

pip install -r requirements.txt

Usage

# Run full pipeline
python main.py

# Run with custom config
python main.py --data_dir data/raw --colormap JET --model rf

Dataset

Place FLIR thermal images (grayscale .jpg / .png) inside data/raw/. Organize into sub-folders by temperature category if you want supervised classification:

data/raw/
├── low/
├── medium/
└── high/

Outputs

Output	Location
Pseudo-colored images	outputs/images/
Grayscale vs color comparison plots	outputs/plots/
Accuracy / classification report	outputs/metrics/
Confusion matrix plot	outputs/plots/
Trained model	models/