tomguluson92 / unsup3d_pytorch3d

(CVPR'20 Best Paper) Unsup3D SoftRas

(WIP)unsup3D_pytorch3d (CVPR2020 Best paper)

Demo | Project Page | Video | Paper

Shangzhe Wu, Christian Rupprecht, Andrea Vedaldi, Visual Geometry Group, University of Oxford. In CVPR 2020 (Best Paper Award).

We propose a method to learn weakly symmetric deformable 3D object categories from raw single-view images, without ground-truth 3D, multiple views, 2D/3D keypoints, prior shape models or any other supervision.

〇 Notice

author: samuel ko

This repo is target to replace the neural_renderer from CVPR 2018 with the sota renderer ICCV 2019 Soft Rasterizer provided by pytorch3d. In order to compatible with the pytorch3d API, we change a little bit in unsup3d/renderer/renderer.py & unsup3d/renderer/utils.py.

However, the effect of this implemention is not good. I will still work on this and try to fix that problem.

SoftRas(trained on celebA)

① Setup (with Anaconda)

1. Install dependencies:

conda env create -f environment.yml

OR manually:

conda install -c conda-forge scikit-image matplotlib opencv moviepy pyyaml tensorboardX

2. Install PyTorch:

conda install pytorch==1.4.0 torchvision cudatoolkit=10.0 -c pytorch

Note: The code is tested with PyTorch 1.4.0 and CUDA 10.0 on Ubuntu 18.04. A GPU version is required for training and testing, since the neural_renderer package only has GPU implementation.

3. Install pytorch3d instead of original repo neural_renderer

conda install --channel https://conda.anaconda.org/pytorch3d pytorch3d

4. (For demo only) Install facenet-pytorch:

This package is optional for the demo. It allows automatic human face detection.

pip install facenet-pytorch

② Datasets

1. CelebA face dataset. Please download the original images (img_celeba.7z) from their website and run celeba_crop.py in data/ to crop the images.
2. Synthetic face dataset generated using Basel Face Model. This can be downloaded using the script download_synface.sh provided in data/.
3. Cat face dataset composed of Cat Head Dataset and Oxford-IIIT Pet Dataset (license). This can be downloaded using the script download_cat.sh provided in data/.

Please remember to cite the corresponding papers if you use these datasets.

③ Demo

python -m demo.demo --input demo/images/human_face --result demo/results/human_face --checkpoint pretrained/pretrained_celeba/checkpoint030.pth

④ Training and Testing

Check the configuration files in experiments/ and run experiments, eg:

python run.py --config experiments/train_gdh.yml --gpu 0 --num_workers 2

⑤ Citation

@InProceedings{Wu_2020_CVPR,
  author = {Shangzhe Wu and Christian Rupprecht and Andrea Vedaldi},
  title = {Unsupervised Learning of Probably Symmetric Deformable 3D Objects from Images in the Wild},
  booktitle = {CVPR},
  year = {2020}
}

unsup3d_pytorch3d's Issues

Excellent work!

Have you figured out the reason of poor results of PyTorch3D?

which pytorch3d version do U used

Thanks for your great work!
Unfortunately, I met some problems when run the demo.py
when I used the pretrained celeba model from the original unsup3d project and then run the demo of yours, the render is broken

I used pytorch 1.5.0 cuda 10.1 and pytorch3d 0.2.5 or 0.3.0 (both are broken)

May I use the original unsup3d pretrained weights or should I train a new one using this repo

May I use the original unsup3d pretrained weights or should I train a new one using this repo