A Realistic Face-to-Face Conversation System based on Deep Neural Networks
This addresses enhancing user experience in avatar conversations, but it is incremental as it builds on existing methods like GANs and sequence-to-sequence models.
The paper tackled improving avatar-based face-to-face conversation by developing a system using sequence-to-sequence models for listening and speaking and a GAN-based synthesizer to generate realistic facial images, resulting in natural facial reactions and realistic outputs as demonstrated with ESPN data.
To improve the experiences of face-to-face conversation with avatar, this paper presents a novel conversation system. It is composed of two sequence-to-sequence models respectively for listening and speaking and a Generative Adversarial Network (GAN) based realistic avatar synthesizer. The models exploit the facial action and head pose to learn natural human reactions. Based on the models' output, the synthesizer uses the Pixel2Pixel model to generate realistic facial images. To show the improvement of our system, we use a 3D model based avatar driving scheme as a reference. We train and evaluate our neural networks with the data from ESPN shows. Experimental results show that our conversation system can generate natural facial reactions and realistic facial images.