Multiple projects for face emotion recognition:

Project for fetching, training and testing google's dataset https://research.google/tools/datasets/google-facial-expression/
It uses multiprocessing to fetch the dataset faster. I am trying to train siamese network with a triplet loss.
To compute face feature pretrained resnet trained on vggface2 is used pipelined into a densenet and then to a fc layer.

Project for face emotion recognition on dataset affectnet and ck+.