Skip to content

Latest commit

 

History

History
47 lines (30 loc) · 3.54 KB

README.md

File metadata and controls

47 lines (30 loc) · 3.54 KB
Raw

Deep Learning for Expression Recognition in Image Sequences

Facial expressions convey lots of information, which can be used for identifying emotions. These facial expressions vary in time when they are being performed. Recognition of certain emotions is a very challenging task even for people. This thesis consists of using machine learning algorithms for recognizing emotions in image sequences. It uses the state-of-the-art deep learning on collected data for automatic analysis of emotions. Concretely, the thesis presents a comparison of current state-of-the-art learning strategies that can handle spatio-temporal data and adapt classical static approaches to deal with images sequences. Expanded versions of CNN, 3DCNN, and Recurrent approaches are evaluated and compared in two public datasets for universal emotion recognition, where the performances are shown, and pros and cons are discussed.

Built With

  • Keras - Keras is a high-level neural networks API, written in Python
  • Tensorflow - TensorFlow™ is an open source software library for numerical computation using data flow graphs.
  • Theano - Theano is a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently.

Datasets

  • SASE-FE - SASE-FE real and fake emotion database contains videos with six typical emotions depicting a real and a fake emotion.
  • Oulu-Casia - Oulu-CASIA NIR&VIS facial expression database contains videos with the six typical expressions.

Author

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

  • The face frontalization [2] preprocess was performed using Douglas Souza [dougsouza] implementation.
  • The 3D CNN model is based on Alberto Montes [albertomontesg] implementation of C3D model.
  • The CNN model is based on Refik Can Malli [rcmalli] implementation of the VGG-Face.
  • The VGG-Face was first introduced by Omkar M. Parkhi, Andrea Vedaldi, Andrew Zisserman from University of Oxford.
  • The C3D Model first introduced by Du Tran, Lubomir Bourdev, Rob Fergus, Lorenzo Torresani, Manohar Paluri from Facebook AI Research and Dartmouth College.

Bibliography

  • [1] Ofodile, I., Kulkarni, K., Corneanu, C. A., Escalera, S., Baro, X., Hyniewska, S., ... & Anbarjafari, G. (2017). Automatic recognition of deceptive facial expressions of emotion. arXiv preprint arXiv:1707.04061.
  • [2] Hassner, T., Harel, S., Paz, E., & Enbar, R. (2015). Effective face frontalization in unconstrained images. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 4295-4304).