RawHandler 0.0.3

A basic library to handle camera raw files for use in machine learning. Built on rawpy and cv2.

RawHandler

RawHandler is a lightweight wrapper around rawpy that provides convenient tools for working with raw sensor data, particularly for training neural networks on raw images.

---

Features

RawHandler can:

1. Open and convert most camera raw files into numpy arrays.

2. Apply black and white point correction automatically.

3. Provide multiple representations of the underlying sensor data:

   • Mono Bayer representation
   • 3-channel sparse representation
   • 4-channel RGGB representation

4. Demosaic Bayer data using colour-demosaicing, supporting:

   • Bilinear interpolation
   • Malvar–He–Cutler (2004)
   • DDFAPD – Menon et al. (2007)

5. Convert color spaces from the camera’s native space to standard targets such as XYZ, sRGB, AdobeRGB, or linear Rec.2020 — all available for every representation.

6. Crop, resize, and generate thumbnails while preserving Bayer pattern alignment.

7. Read EXIF/metadata information (ISO, shutter speed, orientation, etc.) and return it as a convenient Python dictionary.

Currently supported: Bayer raw images In progress: Fujifilm X-Trans support

---

Installation

You can install RawHandler directly from PyPI:

pip install RawHandler

Or install locally from source:

# Clone the repository
git clone https://github.com/rymuelle/RawHandler.git
cd RawHandler

# Editable/development install
pip install -e .

# Standard local install
pip install .

---

Example

A simple demo notebook is available:

examples/simple_demosaicing.ipynb

This example downloads a raw image and demonstrates the basic functionality of RawHandler.

---

License

This project is released under the MIT License.

---

Acknowledgments

Special thanks to the authors of RawNIND:

Brummer, Benoit; De Vleeschouwer, Christophe, 2025. Raw Natural Image Noise Dataset. https://doi.org/10.14428/DVN/DEQCIM, Open Data @ UCLouvain, V1.

---

Updating

1. Update version in pyproject.toml

2. Clean old builds

rm -rf dist build *.egg-info

3. Build

python -m build

4. Check

twine check dist/*

5. Upload

twine upload --repository testpypi dist/*