Group Affect Prediction Using Emotion Heatmaps and Scene Information

In this paper, we describe our work on emotion detection for a group of people in an image. We use an ensemble of a Convolutional Neutral Network (CNN) trained on the emotion heatmaps and a fine-tuned CNN which had been trained on ImageNet data. This work was done for Emotion Recognition in in the Wild (EmotiW) challenge, 2017. Our best submission achieved a test accuracy of 70.64% which was 17.02% above the baseline accuracy provided by the organizers.