navigator8972/vae_assoc

Associative Variational Auto-encoders

vae_assoc

Variational Auto-encoders factored with enforced identity of the latent variables. See reference:

Yin, H., Melo, F. S., Billard, A. and Paiva, A. Associate Latent Encodings in Learning from Demonstrations. In Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence (AAAI-17). San Francisco, California, USA.

Downloading data and a model example

$ cd bin
$ ./download_data.sh
$ cd ../output
$ ./download_model.sh

Video demo

Observe associated encodings of letter images and dynamical writing motion

$ python vae_assoc_model_viewer.py

Inference of writing motion from a letter image

Launch roscore for exchanging messages.

$ roscore

In another terminal, start the the GUI and the rendering simulator

$ python baxter_openrave.py &
$ python baxter_vae_assoc_writer.py

Some main dependencies

Mandatory:

Tensorflow (>=0.8.0)

Optional (for robot motion, messaging and rendering purposes):

PyQT

ROS

OpenRAVE

baxter_pykdl

vae_assoc's Issues

something about your this repository

Hello, today I tried to run your code, but I got an error. Then I found the reason was in your code "baxter_writer.py", you used the function 'baxter_pykdl.baxter_pykdl.baxter_dynamics'. I read the raw code from baxter_pykdl, and there is only baxter_kinetics but not baxter_dynamics. So may I ask for your help that whether you can share your revised baxter_pykdl with me? And my email is [email protected]. Thanks a lot in advance!