notebooks

TorchFSDD Example

This directory contains a Jupyter Notebook demonstration showing how TorchFSDD can be used to quickly generate PyTorch datasets and data loaders, and use these to train a deep recurrent neural network.

The rendered notebook can be viewed here.

Running the notebook

To run the notebook yourself, you will have to first install a number of dependencies.
These are found in requirements.txt.

You should run the following command in the root directory of the repository to install them.

pip install -r notebooks/requirements.txt

Once these are installed, you can run the demo.ipynb notebook using Jupyter Notebook.

jupyter notebook notebooks/demo.ipynb

Model

The particular network that we use is a PyTorch implementation of the DeepGRU[1] architecture, found in model.py.

References

[1]	Mehran Maghoumi & Joseph J. LaViola Jr. "DeepGRU: Deep Gesture Recognition Utility" Advances in Visual Computing, 14th International Symposium on Visual Computing, ISVC 2019, Proceedings, Part I, 16-31.