Pixel3DMM proposes a FLAME fitting optimization that solves for the 3DMM parameters from the uv-coordinate and normal estimates. This approach enables the reconstruction of posed and neutral facial geometry. The model is evaluated on a new benchmark for single-image face reconstruction, which features high diversity facial expressions, viewing angles, and ethnicities. The results show that Pixel3DMM outperforms the most competitive baselines by over 15% in terms of geometric accuracy for posed facial expressions.
Pixel3DMM has a wide range of applications in computer vision, graphics, and face analysis. Its ability to reconstruct 3D faces from single images makes it a valuable tool for face recognition, facial expression analysis, and face synthesis. The model's accuracy and robustness also make it suitable for use in various industries, such as entertainment, healthcare, and security. Additionally, Pixel3DMM's surface normal estimation and uv-coordinate prediction capabilities can be used for other tasks, such as 3D reconstruction and tracking.
