FP-Age: Leveraging Face Parsing Attention for Facial Age Estimation in the Wild

Official repo for our paper FP-Age: Leveraging Face Parsing Attention for Facial Age Estimation in the Wild, accepted by IEEE Transactions on Image Processing (TIP). For IMDB-Clean, please visit https://github.com/yiminglin-ai/imdb-clean.

How to Install

Dependencies

• Numpy
• OpenCV
• PyTorch
• ibug.roi_tanh_warping: See this repository for details: https://github.com/ibug-group/roi_tanh_warping.
• ibug.face_parsing: See this repository for details: https://github.com/hhj1897/face_parsing.
• ibug.face_detection (only needed by the test script): See this repository for details.

After installing the needed dependencies, run the following commands in bash:

git clone --recursive https://github.com/ibug-group/fpage
cd fpage
pip install -e .

Pretrained weights

Download the pretrained weights and unzip them in ibug/age_estimation/weights/

• weights for FP-Age: https://ibug.doc.ic.ac.uk/download/fpage/fpage-resnet50-fcn-14-97.torch
• weights for baselines: https://ibug.doc.ic.ac.uk/download/fpage/baselines.zip

How to run the demo

To run FP-Age using the webcam as input:

python fpage_test.py -i 0 -d cpu # CAUTION: running on CPU is slow

To run on CUDA (Recommended):

python fpage_test.py -i 0 -d cuda:0 

Command-line arguments:

-i VIDEO: Index of the webcam to use (start from 0) or
          path of the input video file
-d: Device to be used by PyTorch (default=cuda:0)

Label map for parsing:

0 : background
1 : skin (including face and scalp)
2 : left_eyebrow
3 : right_eyebrow
4 : left_eye
5 : right_eye
6 : nose
7 : upper_lip
8 : inner_mouth
9 : lower_lip
10 : hair
11 : left_ear
12 : right_ear
13 : glasses

IMDB-Clean

We provide IMDB-Clean for training and testing age estimation models. Please refer to https://github.com/yiminglin-ai/imdb-clean for details.

Baselines

To test baselines, download checkpoints and unzip it under weights folder, then run baseline_test.py:

python baseline_test.py -d cpu --loss dex --age-classes 101
python baseline_test.py -d cpu --loss mv --age-classes 101 # mean variance loss
python baseline_test.py -d cpu --loss dldl --age-classes 101
python baseline_test.py -d cpu --loss dldlv2 --age-classes 101
python baseline_test.py -d cpu --loss ord --age-classes 101

Please refer to our paper for the details of the baselines.

Demo

Reference

Note: If you use this repository in your research, we kindly request you to cite the following paper:

• Lin, Yiming, Jie Shen, Yujiang Wang, and Maja Pantic. "FP-Age: Leveraging Face Parsing Attention for Facial Age Estimation in the Wild." arXiv preprint arXiv:2106.11145 (2021).

Bibtex:

@misc{lin2021fpage,
      title={FP-Age: Leveraging Face Parsing Attention for Facial Age Estimation in the Wild}, 
      author={Yiming Lin and Jie Shen and Yujiang Wang and Maja Pantic},
      year={2021},
      eprint={2106.11145},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}