Code

We often publically release the code developed in-house. Below are a list of such codes released for public use, their short description and their link.

SOFEA: A Non-iterative and Robust Optical Flow Estimation Algorithm for Dynamic Vision Sensors

Repository link: https://bitbucket.org/wengfei_nus/sofea/src/master/

This repository contains a MATLAB implementation of the Single-shot Optical Flow Estimation Algorithm (SOFEA), as presented in the CVPR 2020 workshop paper "SOFEA: A Non-iterative and Robust Optical Flow Estimation Algorithm for Dynamic Vision".

DART: Distribution Aware Retinal Transform

Repository link: Source code (with tracking annotations)

DART is an event-based feature descriptor for silicon retinas. The source code is also part of the submission to a journal, available publicly on arXiV. The DART descriptor has been applied to four different problems, namely object classification, tracking, detection and feature matching.

SLAYER : Spike LAYer Error Reassignment

Repository link: https://bitbucket.org/bamsumit/slayer

SLAYER is an effective backpropagation algorithm for Deep Spiking Neural Networks. This repo is the SLAYER framework implementation that utilizes GPU parallel execution and can be used to train your Spiking Neural Networks on your own data.

SLAYER PyTorch

Repository link: https://github.com/bamsumit/slayerPytorch

This is PyTorch port of SLAYER algorithm for Deep Spiking Neural Networks. Some of the features are still in development.

Spike Files Read Write

Repository link: https://bitbucket.org/bamsumit/spikefilesreadwrite

This repository contains MATLAB scripts to read and write spike data into address event encoded binary format. The binary files are compatible to use with SLAYER.

MATLAB AER Vision Functions

Repository link: https://github.com/gorchard/Matlab_AER_vision_functions

This repository contains some commonly used MATLAB functions for working with and displaying AER vision data.

Neuromorphic Datasets

Link:: https://www.garrickorchard.com/datasets

These are neuromorphic vision datasets captured using DVS sensor: N-MNIST, N-Caltech101, Characters.