TinyTCN by Zanghieri et al. (UNDER REVIEW)

This repository implements the Temporal Convolutional Network (TCN) [1] [2] presented in a manuscript by M. Zanghieri et al. [3] currently under review at the IEEE Access journal. We provide this as supplementary documentation for the peer review now in progress. This small project is a preview of the TCN implementation; upon acceptance of the paper, we will publish the whole curated dataset and code developed for the project. For a technical report about an earlier stage of the same research project, please refer to F. Conti et al. [4].

Usage

To run this small project, clone this repository:

git clone git@github.com:pulp-bio/tcn_under_review.git

The files expose the TCN's implementation and PyTorch and the file tcn_table.txt already contains the generated TCN's structure table.

The requirements (see requirements.txt) are the Python packages PyTorch 1.9.0 and torchinfo 1.8.0, quickly installable via the shell with

python -m pip install -r requirements.txt

The TCN is implemented in the module tcn.py.

1. Run visualize_tcn_table.ipynb (or equivalently visualize_tcn_table.py) to generate the TCN, its torchinfo.ModelStatistics, and the printed table.
2. See the printed table in the standard output or in the file tcn_table.txt.

The generated output file is identical to the one already available in the repository.

Authors

The manuscript documented by this repository involves several authors. The piece of work reported here was developed at the Energy-Efficient Embedded Systems (EEES) Lab of University of Bologna (Italy) by:

• Marcello Zanghieri (Conceptualization, Software, Analysis)
• Prof. Francesco Conti (Supervision, Funding acquisition)
• Prof. Luca Benini (Supervision, Funding acquisition)

Prof. Luca Benini is also with the ETH Zürich (Switzerland).

Citation

@article{key ,
    author = {Zanghieri, M. and others},
    title = {{MANUSCRIPT UNDER REVIEW}},
    journal = {IEEE Access},
    volume = {-},
    year = {2024},
    number = {-},
    pages = {--}
}

References

[1] C. Lea et al., "Temporal convolutional networks for action segmentation and detection," CoRR, vol. abs/1611.05267, 2016.

[2] S. Bai et al., "An empirical evaluation of generic convolutional and recurrent networks for sequence modeling," CoRR, vol. [abs/1803.01271 (https://doi.org/10.48550/arXiv.1803.01271), 2018.

[3] M. Zanghieri et al., MANUSCRIPT UNDER REVIEW, IEEE Access.

[4] F. Conti et al., "AI-powered collision avoidance safety system for industrial woodworking machinery," in AI4DI – Applications. River Publishers, 2021. DOI: 10.1201/9781003337232-17.

License

All files are released under the LGPL-2.1 license (LGPL-2.1) (see LICENSE).