Zhe Wu

Yes. That is what I think. I use the enc_input to train and it seems that the model converged faster. But it is interesting that using enc_output the model also...

The train videos are from the dataset in the work Deception "Detection using Real-life Trial Data (Perez-Rosas et al. 2015)". The original videos can be downloaded in their homepage http://lit.eecs.umich.edu/~deceptiondetection/...