Envy Prediction from Users’ Photos using Convolutional Neural Networks

Envy is often considered a negative trait in human behavior. However, envy also has a positive insight that can motivate a person to accomplish her desired goals. In this paper, we propose a novel method to identify a user’s state of envy (i.e., benign or malicious) based on features from her photos. Specifically, we build a fine-tuned Convolutional Neural Network (CNN) model that takes the user’s photo as input and predicts whether the user has benign or malicious envy characteristics in the given photo. For this study, we create a new dataset containing photos of 255 users of different gender and age group. We conduct ablation studies to build an optimal CNN model to obtain a commendable test accuracy of 97.9%.