Expression Net

Overview:

The Expression-NetTeaser is a deep convolutional neural network (DCNN) model and python code designed for robust estimation of 29 degrees of freedom, 3DMM face expression coefficients from unconstrained face images without the need for face landmark detectors. This project, introduced in the 13th IEEE Conference on Automatic Face and Gesture Recognition in 2018, bundles multiple components for comprehensive 3D face modeling, producing a 3D model mesh file (.ply).

Features:

• Estimating 29D 3DMM face expression coefficients
• Includes 3DMM face identity shape and 6DoF 3D head pose
• Not dependent on fragile landmark detectors, ensuring robustness under various image conditions
• Extremely fast expression estimation
• Provides superior expression estimation compared to methods using landmark detectors

Summary:

The Expression-NetTeaser offers a robust solution for estimating 3DMM face expression coefficients without relying on face landmark detectors. By utilizing deep learning models and python code, it provides fast and accurate expression estimation, making it a valuable tool for 3D face modeling. Users can leverage this project for holistic 3D face modeling, with planned extensions for adding mid-level facial details in the future.