Info: Zenodo’s user support line is staffed on regular business days between Dec 23 and Jan 5. Response times may be slightly longer than normal.

Published August 8, 2024 | Version v1
Dataset Open

Elucidating the Hierarchical Nature of Behavior with Masked Autoencoders

Description

#########

Elucidating the Hierarchical Nature of Behavior with Masked Autoencoders

#########

Authors: Lucas Stoffl, Andy Bonnetto, Stéphane D'Ascoli & Alexander Mathis

Affiliation: Ecole Polytechnique de Lausanne (EPFL)

Date: 25/09/2024

Link to the BiorXiv article : https://doi.org/10.1101/2024.08.06.606796

-----------------

Provided data (hBehaveMAE checkpoints)

We provide a collection of pre-trained models that were reported in our paper, allowing you to reproduce our results for MABe22, hBABEL and Shot7M2 datasets.

Note that you can download Shot7M2 on HuggingFace and generate hBABEL by following the instructions on the github page.

  • hBehaveMAE_hBABEL.pth : checkpoint for the hBehaveMAE pre-trained on the hBABEL dataset
  • hBehaveMAE_Shot7M2.pth : checkpoint for the hBehaveMAE pre-trained on the Shot7M2 dataset
  • hBehaveMAE_MABe22.pth: checkpoint for the hBehaveMAE pre-trained on the MABe22 dataset

References

If you find our code, weights or ideas useful, please cite:

@article {Stoffl2024hBehaveMAE,
    author = {Stoffl, Lucas and Bonnetto, Andy and d{\textquoteright}Ascoli, St{\'e}phane and Mathis, Alexander},
    title = {Elucidating the Hierarchical Nature of Behavior with Masked Autoencoders},
    elocation-id = {2024.08.06.606796},
    year = {2024},
    doi = {10.1101/2024.08.06.606796},
    publisher = {Cold Spring Harbor Laboratory},
    URL = {https://www.biorxiv.org/content/early/2024/08/08/2024.08.06.606796},
    eprint = {https://www.biorxiv.org/content/early/2024/08/08/2024.08.06.606796.full.pdf},
    journal = {bioRxiv}
}

Files

README.md

Files (549.3 MB)

Name Size Download all
md5:514eda25104610be6d4ed8ab7d7c1cf1
188.8 MB Download
md5:84bacf440d0d14294e5d34f7b60ad7ba
244.5 MB Download
md5:dbeb7bc33671d371d42e8f491a4b068d
116.0 MB Download
md5:68b0e48f7fa6a37219f6c6781894ae25
1.8 kB Preview Download

Additional details

Additional titles

Alternative title
BehaveMAE & Shot7M2

Related works

Is part of
Preprint: 10.1101/2024.08.06.606796 (DOI)

Software

Repository URL
https://github.com/amathislab/BehaveMAE
Programming language
Python
Development Status
Active