Pose-disentangled Contrastive Learning for Self-supervised Facial Representation

This repository is the Pytorch implementation for our CVPR2023 paper: Pose-disentangled Contrastive Learning for Self-supervised Facial Representation.

paper link: arxiv

0. Contents

1. Requirements
2. Data Preparation
3. Pre-trained Models
4. Training
5. Evaluation

1. Requirements

To install requirements: Python Version: 3.7.9

pip install -r requirements.txt

2. Data Preparation

You need to download the related datasets and put in the folder which namely dataset.

3. Pre-trained Models

You can download our trained models from Baidu Drive (2qia) and Google Drive .

4. Training

To train the model in the paper, run this command:

python main.py --config_file configs/remote_PCL_vox.yaml

5. Evaluation

We used the linear evaluation protocol for evaluation.

5.1 FER

To evaluate on RAF-DB, run:

python main.py --config_file configs/remote_PCL_linear_eval.yaml

5.2 Pose regression

To trained on 300W-LP and evaluated on AFLW2000, run:

python main_pose.py --config_file configs/remote_PCL_linear_eval_pose.yaml

5.3 Visualization

To visualize on RAF-DB, run:

python visualize.py

TODO

• Refactor the codes of AU detection and face recognition.

IF YOU HAVE ANY PROBLEM, PLEASE CONTACT wangwenbin@cug.edu.cn OR COMMIT ISSUES