Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.
examples added baseline tests Mar 17, 2018
images Add files via upload Mar 16, 2018
sensenet added render mode with rgb_array Mar 20, 2018
tests tests for classification Mar 5, 2018
.gitignore ignore Dec 3, 2017
Dockerfile dockerfile Dec 1, 2017
LICENSE turning into a module WIP Dec 3, 2017 pip works! Dec 29, 2017 Update Mar 16, 2018
Vagrantfile fixed image Nov 4, 2017


SenseNet is a sensorimotor and touch simulator to teach AIs how to interact with their environments via sensorimotor systems and touch neurons. SenseNet is meant as a research framework for machine learning researchers and theoretical computational neuroscientists.


Reinforcement learning

SenseNet can be used in reinforcement learning environments. The original code used OpenAI's gym as the base and so any code written for gym can be used with little to no tweaking of your code. Oftentimes you can just replace gym with sensenet and everything will work. Additionally, SenseNet can be used

Supported Systems

We currently support Mac OS X and Linux (ubuntu 14.04), Windows mostly works, but we don't have a windows developer. We also have docker and vagrant/virtualbox images for you to run an any platform that supports them.

Install from source

git clone you can run "pip install -r requirements.txt" to install all the python software dependencies pip install -e '.[all]'

Install the fast way:

pip install sensenet

Train an basic RL agent to learn to touch a missile with "6th sense":

python examples/agents/ -e TouchWandEnv-v0


I have made and collected thousands of different objects to manipulate in the simulator. You can use the SenseNet dataset or your own dataset. Download the full dataset at



we use pytest to run tests, to tun the tests just type "cd tests && pytest" from the root directory

running benchmarks

Included with SenseNet are several examples for competing on the benchmark "blind object classification" There is a pytorch example and a tensorflow example. to run them: cd agents && python

to see the graphs: tensorboard --logdir runs then go to your browser at http://localhost:6000/ python register sdist upload

You can’t perform that action at this time.