Graph-based Spatial-temporal Feature Learning for Neuromorphic Vision Sensing

Summary

This is the implemtation code and proposed dataset(ASL-DVS) for the following paper. Please cite following paper if you use this code or dataset in your own work. The paper is available via: https://ieeexplore.ieee.org/abstract/document/9199543

MLA:

Bi, Y., Chadha, A., Abbas, A., Bourtsoulatze, E., & Andreopoulos, Y. (2019). Graph-based Spatial-temporal Feature Learning for Neuromorphic Vision Sensing. IEEE Transactions on Image Processing, 9084-9008, 2020

BibTex:

@article{bi2020graph,
 title={Graph-Based Spatio-Temporal Feature Learning for Neuromorphic Vision Sensing},
 author={Bi, Yin and Chadha, Aaron and Abbas, Alhabib and Bourtsoulatze, Eirina and Andreopoulos, Yiannis},
 journal={IEEE Transactions on Image Processing},
 volume={29},
 pages={9084--9098},
 year={2020},
 publisher={IEEE}
 }

Dataset: UCF101-DVS, HMDB-DVS, ASLAN-DVS

We release largest neuromorphic human activity datasets including UCF101-DVS, HMDB-DVS and ASLAN-DVS, and make them available to the research community at the link: https://www.dropbox.com/sh/ie75dn246cacf6n/AACoU-_zkGOAwj51lSCM0JhGa?dl=0

GoogleDrive： HMDB-DVS:https://drive.google.com/file/d/1TVHsmv-Pt2jZVcxfPgpGonXDzqbAAGDB/view?usp=drive_link UCF101-DVS: https://drive.google.com/file/d/1icJFUNJZEofOke-9bizfD-m4643G1ycg/view?usp=drive_link

Code Implementation

Requirements:

 Python 2.7 
 Pytorch 1.0.1.post2
 pytorch_geometric 1.1.2

Preparations:

Training graphs are saved in '../traingraph' folder.
Testing graphs are saved in '../testgraph' folder.
Each sample should contains feature of nodes, edge, pseudo adresses and label.

Running examples:

cd code
python main.py   # running file for RGCNN+Plain 3D 

#The results can be found in the 'Results' folder.

Contact

For any questions or bug reports, please contact Yin Bi at [email protected] .