About

A quick way to start with machine and deep learning¶

News

You can now overwrite configuration with command line options. See Overwriting configuration with CLI!

News

You can now define a custom train/validation split for experiments. See Defining custom splits!

Kit4DL is a Python framework for simple, fast, and customisable deep learning prototyping and development. It uses PyTorch Lightning as a backend, since it is a powerful, yet quick-to-set-up, framework for deep learning pipelines management, and it relies on TOML (TOML format documentation can be found in https://toml.io/) configuration file. Though TOML is a relatively new format, it is gaining more and more attention in the Python ecosystem thanks to its intuitiveness, conciseness, and ease of reading. Furthermore, Kit4DL uses torchmetrics library for PyTorch-integrated and efficient implementation of a miscellaneous of metrics.

🖋️ Authors¶

Jakub Walczak
Marco Macini
Mirko Stojiljkovic
Shahbaz Alvi

🙏 Acknowledgement¶

This work has received fundings from the Polish National Centre for Research and Development under the LIDER XI program [grant number 0092/L-11/2019, "Semantic analysis of 3D point clouds"] and from the European Union’s Horizon 2020 Research and Innovation programme [SILVANUS Project - grant agreement number 101037247].

📜 Cite Us¶

@SOFTWARE{kit4dl,
  author = {Walczak, Jakub and
            Mancini, Marco and
            Stojiljković, Mirko and
            Alvi, Shahbaz},
  title = {Kit4DL},
  month = sep,
  year = 2023,
  note = {{Available in GitHub: https://github.com/opengeokube/kit4dl}},
  publisher = {Zenodo},
  version = {2023.9b1},
  doi = {10.5281/zenodo.8328176},
  url = {https://doi.org/10.5281/zenodo.8328176}
}