Datasets
Code
LAUD is the implementation of our LBP-based AU detector, described in this paper. It is developed as a standalone, WIN32 command-line version of the AU detector implemented in the SEMAINE project. It comes with trained models for the following AUs: AU1, AU2, AU4, AU5, AU6, AU7, AU9, AU12, AU14, AU15, AU20, AU25, AU27, and AU45. The detector does not perform equally well for all these AUs. The performance for AU1, AU2, AU4, AU12, and AU25 was evaluated as being well enough for use in the SEMAINE project (http://semaine.opendfki.de/wiki). The detector assumes that the input image is in frontal view (if the -f argument is passed, face detection but not registration will be carried out).
More information about the point detector can be found in:
B. Jiang, M.F. Valstar and M. Pantic, "Action Unit detection using sparse appearance descriptors in space-time video volumes", in Proc. IEEE Int'l Conf. on Automatic Face and Gesture Recognition (FG'11), 2011.
We kindly request you to cite this work if you decide to use the point detector for research purposes.
Download the LAUD setup.
How to use it: It is possible to detect an AU in a single image:
LAUD C:\LAUD\Laud_SVM_unregistered_UniformLBP_AU1.xml C:\tmp\img0001.png -v -f
or in all images in a directory:
LAUD C:\LAUD\Laud_SVM_unregistered_UniformLBP_AU1.xml C:\tmp\imdir\ -v -f -d
Call LAUD without input arguments to see usage options. If you process a whole directory, the results will be stored in that directory.