Material for paper Facial Expression Recognition using Convolutional Neural Networks: State of the Art. The scripts/
folder contains a script that was used for applying histogram equalization to the FER2013 dataset. The models/
folder contains trained Torch7 models (Table IV in the paper):
File | Architecture | Depth | FER2013 Test Accuracy |
---|---|---|---|
vgg.t7 |
VGG | 10 | 72.7% |
inception.t7 |
Inception/GoogLeNet | 16 | 71.6% |
resnet.t7 |
Resnet-34 | 33 | 72.4% |
Author: Christopher Pramerdorfer, TU Wien
Licence: zlib