Projectsresearch projects, tools, libraries

A python library for machine learning research on motion capture data

Teach a neural network to generate novel, beat-synchronous dance movements for a given song.

Visualizing the learning process of artificial neural networks

Implementations of RBM and its variants in tensorflow

WalkNet is a neural-network-based interactive movement controller with navigation capabilities, as well as modulating the agent's affective state and movement signature.

AffectNet is a generative model of affect-expressive movements, based on FCRBM. The affective qualities are represented along the valence and arousal dimensions.

A custom Keras layer for Mixter Density outputs

MocapJS is a motion capture library for Three.js. It is designed for applications such as web-based mocap players, streaming movement data, and virtual reality scenarios.

Mova is a web-based platform for visualization and analysis of human movement data, based on D3.js.

A javascript port of Factored Conditional Restricted Boltzmann Machines (FCRBM).

This project explores what a deep neural network that is trained on human movement learns about movement.

A small C++ program to convert motion capture marker data in C3D files to CSV format.

AIMS is a framework for agent interaction simulations, developed in Java.

This project implements a fully integrated meteo-ocean forecasting system for Caspian Sea and Persian Gulf.