Spatiotemporal Recurrent Convolutional Networks for Recognizing Spontaneous Micro-Expressions
These codes are used for micro-expression recognition on individual datasets. The methods can be accessed by the paper "Spatiotemporal Recurrent Convolutional Networks for Recognizing Spontaneous Micro-Expressions, IEEE TMM2020", which includes the STRCN-A and STRCN-G.
The code was written in MATLAB R2018b, and tested on Windows 7 and 10.
- MATLAB's Computer Vision Toolbox
- MatConvNet (Beta23)
- Eulerian Video Magnification (2012):http://people.csail.mit.edu/mrub/vidmag/
- The faces should be cropped by your own tools and applified by the EVM.
- All images are storaged with the format of casme2 way.
- The appearance connect and geometric connect needs to be performed before training the deep model. All the data needs to be prepared and storaged into the mat file.