MatcherNet is a probabilistic state-space model for dynamic system identification and control. With MatcherNet, you can easily design a dynamic world model of high-dimensional / multi-modal / multi-scale states and observations for robotics, image processing, sensor networks, and their hybrid cases. MatcherNet may provide you a solution better than a large state space model, a deep neural network of end-to-end structure, etc.

MatcherNet was developped with support by the New Energy and Industrial Technology Development Organization (NEDO), Japan, and by Post-K application development for exploratory challenges from the MEXT, Japan.

Copyright (C) 2019 Kyoto University and National Institute of Advanced Industrial Science and Technology (AIST)

See an overview for knowing what is MatcherNet in a little bit more detail.

See the matchernet documentation for important classes and functions.

See a developers' info. for knowing some additional info. for installation.

The beta version has been released below: https://pypi.org/project/matchernet/

You can install the beta version with the following command:

pip install matchernet

See demo files under the /demos directory.

Python3 demos/demo_ekf.py

See also jupyter notebooks under the /examples directory.

It is tested on 3.7.2 and 3.7.3. 3.8.* is not supported yet.

mkdir ~/virt_env
cd ~/virt_env
virtualenv -p Python3 mn
workon mn

pip3 install -r requirements.txt

export PYTHONPATH="/path/to/dir:$PYTHONPATH"

You need this if you are to modify files in the matchernet original packages.

Python3 demos/demo_ekf.py