/SS-encoder-video

Primary LanguageJupyter NotebookGNU General Public License v3.0GPL-3.0

State-space encoding applied on the video stream output system of the ball in box environment.

The code accompanying a published paper. link

@article{beintema2021non,
  title={Non-linear state-space model identification from video data using deep encoders},
  author={Beintema, Gerben I and Toth, Roland and Schoukens, Maarten},
  journal={IFAC-PapersOnLine},
  volume={54},
  number={7},
  pages={697--701},
  year={2021},
  publisher={Elsevier}
}

The code here has been updated to be compediable with the newest version of deepSI (0.3.0). The exact version used in the paper can be found at 7f375bf4013e294779bebbee28687cb66c45a23d for this repository and 0216a6fe9f86d93fb1296de71af2289295dae1bf in deepSI.

Instructions

  • Get anaconda 3 (python>= 3.6)
  • Install pytorch (Instructions CUDA optional)
  • Install deepSI
    • git clone git@github.com:GerbenBeintema/deepSI.git
    • cd deepSI
    • pip install -e .
  • install jupyter notebook
    • (i.e. conda install -c conda-forge jupyterlab)
  • Use jupyter notebooks to open notebooks.
    • (e.g. jupyter notebook ball-in-box-encoder-approach.ipynb)

State-space encoding structure

encoder image