Here are our main code about EmotiW2018_task1 challenge, In the task1, we win the second place and the details are refered in our paper "Cascade Attention Networks For Group Emotion Recognition with Face, Body and Image Cues. International Conference on Multimodal Interaction (ICMI’18), ACM."
pytorch-0.4.0, mtcnn, openopen, caffe
We only provide the code for feature extraction, training and testing. As our cascade attention networks is still modified for further improvement, we will release the model code soon.
The cascade attention networks, sample code and test code have been unloaded, you can use this code to train a model that use face cues. Other models we use a normal method to train, then use the combine code to get the fusion result.