radekd91/emoca
A CVPR 2022 deep learning method that reconstructs 3D face geometry and expression from a single input image.
Not currently ranked — collecting fresh signals.
star history
EMOCA is a deep learning-based system for monocular face capture and animation. It takes a single face image as input and produces a 3D reconstruction including shape, expression, and detailed geometric displacements. The method uses a deep neural network trained on facial imagery to regress 3D face model parameters, achieving robust reconstruction even for highly emotional expressions captured in the wild. The implementation relies on PyTorch and PyTorch Lightning for model training and inference.
Frequently asked
- What is radekd91/emoca?
- A CVPR 2022 deep learning method that reconstructs 3D face geometry and expression from a single input image.
- Is emoca open source?
- Yes — radekd91/emoca is an open-source project tracked on heatdrop.
- What language is emoca written in?
- radekd91/emoca is primarily written in Python.
- How popular is emoca?
- radekd91/emoca has 850 stars on GitHub.
- Where can I find emoca?
- radekd91/emoca is on GitHub at https://github.com/radekd91/emoca.