/UCVME

AAAI 2023: Semi-Supervised Deep Regression with Uncertainty Consistency and Variational Model Ensembling via Bayesian Neural Networks

Primary LanguagePython

Semi-Supervised Deep Regression with Uncertainty Consistency and Variational Model Ensembling via Bayesian Neural Networks

This is the implementation of UCVME for the paper "Semi-Supervised Deep Regression with Uncertainty Consistency and Variational Model Ensembling via Bayesian Neural Networks".

UCVME



Data

Researchers can get the UTKFace dataset from https://susanqq.github.io/UTKFace/ (Aligned&Cropped Faces). Extract the zip file and set up the files according to the example files in DATA_DIR/UTKFace

DATA_DIR
 |_ FileList.csv
 |_ UTKFace
    |_ 1_0_0_20161219140623097.jpg.chip.jpg
    |_ 1_0_0_20161219140627985.jpg.chip.jpg
    |_ 1_0_0_20161219140642920.jpg.chip.jpg
    ...


Environment

It is recommended to use PyTorch conda environments for running the program. A requirements file has been included.



Training and testing

To perform training, run:

python3 ucvme_age.py --output=<OUTPUT_DIR>

To perform testing only, run:

python3 ucvme_age.py --output=<OUTPUT_DIR> --test_only


Pretrained models

Trained checkpoints and models for the 10% labeled dataset setting can be downloaded from: https://hkustconnect-my.sharepoint.com/:f:/g/personal/wdaiaj_connect_ust_hk/Epq-44XUV_lIoUe7IdkZo44B6vBgiqGIxo6tgCxMQsU48A?e=Uf3GQu

To run with the pretrained model weights, replace the .pts files in the target output directory with the downloaded files.


Experiments MAE R2
10% labeled dataset 5.26 ± 0.02 57.9% ± 0.3


Notes



Citation

If this code is useful for your research, please consider citing:

@article{dai2023semi,
  title={Semi-Supervised Deep Regression with Uncertainty Consistency and Variational Model Ensembling via Bayesian Neural Networks},
  author={Dai, Weihang and Li, Xiaomeng and Cheng, Kwang-Ting},
  journal={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={37},
  number={6},
  pages={7304--7313},
  year={2023}
}