An Efficient Spiking Neural Network for Recognizing Gestures with a DVS Camera on the Loihi Neuromorphic Processor

This repository provides the source codes for analyzing the DNN-to-SNN conversion through SNNToolbox, and the codes for pre-processing the DvsGesture dataset to make it possible to train in the DNN domain. For more details, please refer to our IJCNN '20 paper. If you used these results in your research, please refer to the paper

R. Massa, A. Marchisio, M. Martina and M. Shafique, "An Efficient Spiking Neural Network for Recognizing Gestures with a DVS Camera on the Loihi Neuromorphic Processor," 2020 International Joint Conference on Neural Networks (IJCNN), Glasgow, United Kingdom, 2020, pp. 1-9, doi: 10.1109/IJCNN48605.2020.9207109.

@INPROCEEDINGS{Massa2020AnEfficientSNN,
  author={R. {Massa} and A. {Marchisio} and M. {Martina} and M. {Shafique}},
  booktitle={2020 International Joint Conference on Neural Networks (IJCNN)}, 
  title={An Efficient Spiking Neural Network for Recognizing Gestures with a DVS Camera on the Loihi Neuromorphic Processor}, 
  year={2020},
  volume={},
  number={},
  pages={1-9},
  doi={10.1109/IJCNN48605.2020.9207109}}

Using scripts

The folder SNNToolbox_analysis contains the scripts for analyzing the DNN-to-SNN conversion through SNNToolbox.

• simulator.py requires as input a DNN trained in Keras (in .h5 format), and executes the conversion with the parameters specified in sim_specs.txt.
• sweep_simulator.py executes a sweep of the simulation with different parameters, specified in sim_sweep_specs.txt.

The folder DVS-preprocessing contains the scripts for processing the DvsGesture dataset, and make it possible to be used in the DNN domain.

• dataset_generator{...}.py generates the dataset with the specified settings and exports it in .pickle format.
• dataset_video{...}.py generates the video with the specified settings.
• some examples of output results are available.