deeplabcut2yolo 2.2.8

Convert DeepLabCut dataset to YOLO format

deeplabcut2yolo

Convert DLC to YOLO,
Lightning-fast and hassle-free.

deeplabcut2yolo facilitates training DeepLabCut datasets on YOLO models. Deeplabcut2yolo automatically converts DeepLabCut (DLC) labels to COCO-like format compatible with YOLO, while providing customizability for more advanced users, so you can spend your energy on what matters!

All DeepLabCut datasets belong to their respective owner under CC BY-NC 4.0. This particular image is the training data for YOLO, converted using deeplabcut2yolo from the Tri-Mouse dataset (Lauer et al., 2022).

Quick Start

import deeplabcut2yolo as d2y

# In its simplest form,
d2y.convert("./deeplabcut-dataset/")

# To also generate data.yml
d2y.convert(
    dataset_path,
    train_paths=train_paths,
    val_paths=val_paths,
    skeleton_symmetric_pairs=skeleton_symmetric_pairs,
    data_yml_path="data.yml",
    class_names=class_names,
    verbose=True,
)

To install deeplabcut2yolo using pip:

pip install deeplabcut2yolo

For more information, see examples and documentation.

Features

• Automatically detect default DeepLabCut dataset structure
• Vectorized label conversion
• Support single- and multi-animal projects
• Convenient data.yml generation function for YOLO models

Contribution

You can contribute to deeplabcut2yolo by making pull requests. Currently, these are high-priority features:

• Testing module and test cases
• Documentation

Citation

Citation is not required but is greatly appreciated. If this project helps you, please cite using the following APA-style reference:

Pornsiriprasert, S. (2025). Deeplabcut2yolo: A DeepLabCut-to-YOLO Dataset Converter for Python (v2.2.7). Zenodo. https://doi.org/10.5281/zenodo.17386187

or this BibTeX entry.

@software{pornsiriprasert2025,
  author       = {Pornsiriprasert, Sira},
  title        = {Deeplabcut2yolo: A DeepLabCut-to-YOLO Dataset Converter for Python},
  month        = dec,
  year         = 2025,
  publisher    = {Zenodo},
  version      = {v2.2.7},
  doi          = {10.5281/zenodo.17386187},
  url          = {https://doi.org/10.5281/zenodo.17386187},
}