1arpanet / syna

:revolving_hearts: Emotion Recognition based on Spatio-Temporal Machine Learning.

Emotion Recognition based on Spatio-Temporal Machine Learning. You can find more information about the project in my Master's thesis.

Installation [Ubuntu 16.04 LTE]

1. Run sudo gedit /etc/apt/sources.list and change the line deb http://us.archive.ubuntu.com/ubuntu/ xenial main restricted to deb http://us.archive.ubuntu.com/ubuntu/ xenial main universe

2. Run install.sh

Repository Structure

• /data contains auxiliary data for training and processing
• /datasets is expected to contain the datasets necessary for training syna. The datasets so far are The Extended Cohn-Kanade (CK+) database and Acted Facial Expressions In The Wild database 6.0. Given that I have no rights to publish those datasets, you'll have to contact the authors for download
• /experiments contains a set of experiments to test the developed models, together with a set of utils
• /scripts contains useful scripts to parse the datasets before applying the syna and deepsyna models. Make sure that the datasets are contained in the datasets directory before executing
• /syna contains the main codebase of the project: C3D, dataloaders, face-frontalization, a custom version of OpenFace, and LSTM.

Citation

If you use any of the resources provided on this page in any of your publications, I ask you to cite the following work:

Syna: Emotion Recognition based on Spatio-Temporal Machine Learning Daniyal Shahrokhian, KTH Royal Institute of Technology, 2017

Open Source Acknowledgements

OpenFace

C3D Keras

Face Frontalization

References

Constrained Local Neural Fields for robust facial landmark detection in the wild

Learning Spatiotemporal Features with 3D Convolutional Networks

Effective Face Frontalization in Unconstrained Images