Garment Pose Estimation with Differentiable Rendering
Differentiable Garment Pose is an advanced framework for estimating garment poses by leveraging the power of differentiable rendering. This project combines computer vision, graphics, and machine learning to create a pipeline capable of understanding and predicting garment deformations in 3D.
In progress features:
- use a config file to set up the model
- currently only works for a single set of garment and pose. provide potential support for multiple garments and poses
- fix the hyperparameter optimization script
- repo cleanup and streamline generated output
- build a docker container
- Differentiable Rendering: Smooth and continuous optimization of garment poses using rendering gradients.
- Modular Design: Easily integrate and customize for various garment types and pose scenarios.
- GPU accelerated
- Python >= 3.8
- PyTorch >= 1.12
- CUDA support (optional for GPU acceleration)
- SMPL-X model (download from here)
Coming soon!