unsupervised-detail-layer

Code repository for paper "Self-supervised Learning of Detailed 3D Face Reconstruction (TIP2020)"

Prerequisite

python 3.7
tensorflow 1.15
g++ 4.8 (other version might also work)
tf_mesh_renderer

Please read install.sh for more details. It installs all packages required to run the codes, and compiles the c++ kernel of the differentiable renderer. Pay attention to finding the correct path to TF_INC and TF_LIB. The path specified in install.sh might not suit your case. If it does not work, please try to find them manually. You can also compile the codes using methods provided by tf_mesh_renderer.

Files

We release the checkpoint files, resources, and some testing results for our model. As the training tfrecords are huge, we only provide one tfrecord for testing our codes. We will release the url in baidu drive in some days.

google drive

Please download the training data train_data.zip into the repo root directory ./ and unzip them.
Please download the checkpoints ckptxxx.zip into ./results/ and unzip them.
Please download the resources.zip into the repo root directory ./ and unzip it.
Please download the frozen_models.zip into ./preprocess/ and unzip it. They are face detection and landmark detection models used in proprocessing a face image.

Running

In run_train.sh, we provide commands for training the coarse / fine model.
In run_test.sh, we provide commands for inference. Note that the batch_size should be 1.

Citation

If you use the codes provided here in your paper, please cite the following.

@article{chen2019self,
  title={Self-supervised Learning of Detailed 3D Face Reconstruction},
  author={Chen, Yajing and Wu, Fanzi and Wang, Zeyu and Song, Yibing and Ling, Yonggen and Bao, Linchao},
  journal={IEEE Transactions on Image Processing},
  year={2020}
}

Acknowledgement

Our codes use the differentiable renderer and vggface backbone provided in the following two repos. We thank them for providing the codes.

tf_mesh_renderer: https://github.com/google/tf_mesh_renderer
vgg face: https://github.com/ZZUTK/Tensorflow-VGG-face

Contact

If you have any questions regarding this work, please send emails to jadechancyj907@gmail.com.

License

MIT License

Copyright (c) [2020] [Yajing Chen]

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

cyj907/unsupervised-detail-layer