Landmark Guided 4D Facial Expression Generation

TL;DR

This paper introduces LM-4DGAN, a generative model that synthesizes 4D facial expressions guided by neutral landmarks, improving identity robustness through novel discriminator, autoencoder, and attention mechanisms.

Contribution

The paper presents a landmark-guided 4D facial expression generation model with enhanced identity robustness using a discriminator, autoencoder, and cross-attention.

Findings

Improved identity robustness in 4D facial expression synthesis.

Effective use of neutral landmarks for expression guidance.

Enhanced generation quality with attention mechanisms.

Abstract

In this paper, we proposed a generative model that learns to synthesize the 4D facial expression with the neutral landmark. Existing works mainly focus on the generation of sequences guided by expression labels, speech, etc, while they are not robust to the change of different identities. Our LM-4DGAN utilizes neutral landmarks to guide the facial expression generation while adding an identity discriminator and a landmark autoencoder to the basic WGAN for achieving better identity robustness. Furthermore, we add a cross-attention mechanism to the existing displacement decoder which is suitable for the given identity.