A Novel Multi-Attribute Face-to-Cartoon Model for Human-Computer Interaction

Human-computer interaction in social media have attracted increasing attentions from both academic and industry. To explore a new engaging and interesting form of human-computer interaction, this paper introduces a novel multi-attribute face-to-cartoon system for social media. This system can generate a corresponding emoji from a face snapshot. The system first captures the target face of the selected snapshot and then utilizes a convolutional network to predict its attribute, such as emotion, gender, and glasses. By using this attribute information, this system aims at finding a corresponding emoji. Furthermore, a demo application based on Ubuntul6.04 is also introduced. The experimental results with this demo application verify the performance of the proposed method.