Face Alignment in Full Pose Range: A 3D Total Solution

This project is authored by Jianzhu Guo, Xiangyu Zhu and partially supervised by Zhen Lei.

[Updates]

• 2018.12.23: Add several features: depth image estimation, PNCC, PAF feature and obj serialization. See dump_depth, dump_pncc, dump_paf, dump_obj options for more details.
• 2018.12.2: Support landmark-free face cropping, see dlib_landmark option.
• 2018.12.1: Refine code and add pose estimation feature, see utils/estimate_pose.py for more details.
• 2018.11.17: Refine code and map the 3d vertex to original image space.
• 2018.11.11: Update end-to-end inference pipeline: infer/serialize 3D face shape and 68 landmarks given one arbitrary image, please see readme.md below for more details.
• 2018.10.4: Add Matlab face mesh rendering demo in visualize.
• 2018.9.9: Add pre-process of face cropping in benchmark.

[Todo]

• Depth image estimation.
• PNCC (Projected Normalized Coordinate Code).
• PAF (Pose Adaptive Feature).
• Obj serialization with sampled texture.
• Update the face detector (not Dlib).
• Training details.
• Face swapping.
• Face Profiling.

Introduction

This repo holds the pytorch improved re-implementation of paper Face Alignment in Full Pose Range: A 3D Total Solution. Several external works are added in this repo, including real-time training, training strategy and so on. Therefore, this repo is far more than re-implementation. One related blog will be published for some important technique details in future. As far, this repo releases the pre-trained first-stage pytorch models of MobileNet-V1 structure, the pre-processed training&testing dataset and codebase. It is worth to note that the inference time is about 0.27ms per image (input batch with 128 images) on GeForce GTX TITAN X. Why not evaluate it on single image? Because most time for single image is spent on function call. The inference speed is equal to MobileNet-V1 with 120x120x3 tensor as input, therefore it is possible to convert to mobile devices.

This repo will keep updating in my spare time, and any meaningful issues and PR are welcomed.

Several results on ALFW-2000 dataset (inferenced from model phase1_wpdc_vdc.pth.tar) are shown below.

Applications & Features

1. Face Alignment

2. Face Reconstruction

3. 3D Pose Estimation

4. Depth Image Estimation

5. PNCC & PAF Features

Getting started

Requirements

• PyTorch >= 0.4.1
• Python >= 3.6 (Numpy, Scipy, Matplotlib)
• Dlib (Dlib is optionally for face and landmarks detection. There is no need to use Dlib if you can provide face bouding bbox and landmarks. Besides, you can try the two-step inference strategy without initialized landmarks.)
• OpenCV (Python version, for image IO opertations.)
• Cython (For accelerating depth and PNCC render.)
• Platform: Linux or macOS (Windows is not tested.)

# installation structions
sudo pip3 install torch torchvision # for cpu version. more option to see https://pytorch.org
sudo pip3 install numpy scipy matplotlib
sudo pip3 install dlib==19.5.0 # 19.15+ version may cause conflict with pytorch in Linux, this may take several minutes
sudo pip3 install opencv-python
sudo pip3 install cython

In addition, I strongly recommend using Python3.6+ instead of older version for its better design.

Usage

1. Clone this repo (this may take some time as it is a little big)

   git clone https://github.com/cleardusk/3DDFA.git  # or git@github.com:cleardusk/3DDFA.git
   cd 3DDFA
   

   Then, download dlib landmark pre-trained model in Google Drive or Baidu Yun, and put it into models directory. (To reduce this repo's size, I remove some large size binary files including this model, so you should download it : ) )

2. Build cython module (just one line for building)

   cd utils/cython
   python3 setup.py build_ext -i
   

   This is for accelerating depth estimation and PNCC render since Python is too slow in for loop.

3. Run the main.py with arbitrary image as input

   python3 main.py -f samples/test1.jpg
   

   If you can see these output log in terminal, you run it successfully.

   Dump tp samples/test1_0.ply
   Save 68 3d landmarks to samples/test1_0.txt
   Dump obj with sampled texture to samples/test1_0.obj
   Dump tp samples/test1_1.ply
   Save 68 3d landmarks to samples/test1_1.txt
   Dump obj with sampled texture to samples/test1_1.obj
   Dump to samples/test1_pose.jpg
   Dump to samples/test1_depth.png
   Dump to samples/test1_pncc.png
   Save visualization result to samples/test1_3DDFA.jpg
   

   Because test1.jpg has two faces, there are two .ply and .obj files (can be rendered by Meshlab or Microsoft 3D Builder) predicted. Depth, PNCC, PAF and pose estimation are all set true by default. Please run python3 main.py -h or review the code for more details.

   The 68 landmarks visualization result samples/test1_3DDFA.jpg and pose estimation result samples/test1_pose.jpg are shown below:

4. Additional example

   python3 ./main.py -f samples/emma_input.jpg --bbox_init=two --dlib_bbox=false
   

Citation

@article{zhu2017face,
  title={Face Alignment in Full Pose Range: A 3D Total Solution},
  author={Zhu, Xiangyu and Lei, Zhen and Li, Stan Z and others},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2017},
  publisher={IEEE}
}

@misc{3ddfa_cleardusk,
  author =       {Jianzhu Guo, Xiangyu Zhu and Zhen Lei},
  title =        {3DDFA},
  howpublished = {\url{https://github.com/cleardusk/3DDFA}},
  year =         {2018}
}

Inference speed

When input batch size is 128, the total inference time of MobileNet-V1 takes about 34.7ms. The average speed is about 0.27ms/pic.

Evaluation

First, you should download the cropped testset ALFW and ALFW-2000-3D in test.data.zip, then unzip it and put it in the root directory. Next, run the benchmark code by providing trained model path. I have already provided five pre-trained models in models directory (seen in below table). These models are trained using different loss in the first stage. The model size is about 13M due to the high efficiency of MobileNet-V1 structure.

python3 ./benchmark.py -c models/phase1_wpdc_vdc.pth.tar

The performances of pre-trained models are shown below. In the first stage, the effectiveness of different loss is in order: WPDC > VDC > PDC. While the strategy using VDC to finetune WPDC achieves the best result.

Model	AFLW (21 pts)	AFLW 2000-3D (68 pts)	Download Link
phase1_pdc.pth.tar	6.956±0.981	5.644±1.323	Baidu Yun or Google Drive
phase1_vdc.pth.tar	6.717±0.924	5.030±1.044	Baidu Yun or Google Drive
phase1_wpdc.pth.tar	6.348±0.929	4.759±0.996	Baidu Yun or Google Drive
phase1_wpdc_vdc.pth.tar	5.401±0.754	4.252±0.976	In this repo.

Training

The training scripts lie in training directory. The related resources are in below table.

Data	Download Link	Description
train.configs	BaiduYun or Google Drive, 217M	The directory contraining 3DMM params and filelists of training dataset
train_aug_120x120.zip	BaiduYun or Google Drive, 2.15G	The cropped images of augmentation training dataset
test.data.zip	BaiduYun or Google Drive, 151M	The cropped images of AFLW and ALFW-2000-3D testset

After preparing the training dataset and configuration files, go into training directory and run the bash scripts to train. The training configutations are all presented in bash scripts.

FQA

1. Face bounding box initialization

   The original paper shows that using detected bounding box instead of ground truth box will cause a little performance drop. Thus the current face cropping method is robustest. Quantitative results are shown in below table.

2. Face reconstruction

   The texture of non-visible area is distorted due to self-occlusion, therefore the non-visible face region may appear strange (a little horrible).

Acknowledgement

• Thanks for Yao Feng's fantastic works PRNet and face3d.
• Thanks for the official PyTorch twitter account's tweet.

Thanks for your interest in this repo. If your work or research benefit from this repo, please cite it, star it and popularize it 😃

Welcome to focus on my 3D face related works: MeGlass and Face Anti-Spoofing.

Contact

Jianzhu Guo (郭建珠) [Homepage, Google Scholar]: jianzhu.guo@nlpr.ia.ac.cn.