solangii / PhyDIR

PhyDIR(2022 CVPR) Unofficial PyTorch Implementation

PhyDIR

Pytorch implementation of Physically-Guided Disentangled Implicit Rendering for 3D Face Modeling (2022 CVPR)

💻 Setup

env: nsml v2
gpu: V100-1

1. Environment setup
   • set up your environment following the Unsup3D's guide
   • OR use env.yaml
2. Dataset setup
3. Training
   • change phydir path, conda path and config name

Example:

cd [phydir] && \
source /home/nsml/anaconda3/etc/profile.d/conda.sh && \
conda activate unsup3d && \
python run.py --config configs/[cfg_name].yml --gpu 0 --num_workers 4

📁 File Structure

--------------------
├── data
│   ├── celeba
│   │   ├── train
│   │   ├── val
│   │   ├── test
│   │   │   ├── datalist ([id].json)
│   │   │   ├── *.jpg
|   ├── casia
|   ├── celebamask_hq
├── configs
│   ├── *.yml
├── phydir
│   ├── datasets
|   ├── models
|   ├── *.py
├── results
├── run.py 
--------------------

---

References

• Physically-guided Disentangled Implicit Rendering for 3D Face Modeling
• Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild code
• Learning to Aggregate and Personalize 3D Face from In-the-Wild Photo Collection