Training deep models for emotion recognition using preprocesing kernels and different architectures.
Autoencoder.py: Train an autoencoder architecture to create a high dimensional space embbeding for futher perform emotion classifier.
MiniXceptionTrain.py: Train a small version of the Xception architecture composed with separable convolutions and residual modules to perform emotion recognition in the FER2013 low resolution data base.
facenet.py: Finetunning of a FaceNet Inception-Resnet-v1 architeture to perform emotion classification on different databases.
grimmanet.py: Implementation of novel Inception modules to perform robust emotion recognition.
Train_CNN.py: Transfer learning and finetunning of state of art deep convolutional models (Google Inception v3 and MobileNet v2) to perform facial expression classification.