Home > Research > Publications & Outputs > Reinforced Learning for Label-Efficient 3D Face...

Electronic data

  • ICRA2023

    Accepted author manuscript, 6.47 MB, PDF document

    Available under license: CC BY: Creative Commons Attribution 4.0 International License

View graph of relations

Reinforced Learning for Label-Efficient 3D Face Reconstruction

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Forthcoming

Standard

Reinforced Learning for Label-Efficient 3D Face Reconstruction. / Mohaghegh, Hoda; Rahmani, Hossein; Laga, Hamid et al.
IEEE International Conference on Robotics and Automation (ICRA). 2023.

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Harvard

Mohaghegh, H, Rahmani, H, Laga, H, Boussaid, F & Bennamoun, M 2023, Reinforced Learning for Label-Efficient 3D Face Reconstruction. in IEEE International Conference on Robotics and Automation (ICRA).

APA

Mohaghegh, H., Rahmani, H., Laga, H., Boussaid, F., & Bennamoun, M. (in press). Reinforced Learning for Label-Efficient 3D Face Reconstruction. In IEEE International Conference on Robotics and Automation (ICRA)

Vancouver

Mohaghegh H, Rahmani H, Laga H, Boussaid F, Bennamoun M. Reinforced Learning for Label-Efficient 3D Face Reconstruction. In IEEE International Conference on Robotics and Automation (ICRA). 2023

Author

Mohaghegh, Hoda ; Rahmani, Hossein ; Laga, Hamid et al. / Reinforced Learning for Label-Efficient 3D Face Reconstruction. IEEE International Conference on Robotics and Automation (ICRA). 2023.

Bibtex

@inproceedings{9de6dadbc0af489895400fbe84ad5eba,
title = "Reinforced Learning for Label-Efficient 3D Face Reconstruction",
abstract = "3D face reconstruction plays a major role in many human-robot interaction systems, from automatic face authentication to human-computer interface-based entertainment. To improve robustness against occlusions and noise, 3D face reconstruction networks are often trained on a set of in-the-wild face images preferably captured along different viewpoints of the subject. However, collecting the required large amounts of 3D annotated face data is expensive and time-consuming. To address the high annotation cost and due to the importance of training on a useful set, we propose an Active Learning (AL) framework that actively selects the most informative and representative samples to be labeled. To the best of our knowledge, this paper is the first work on tackling active learning for 3D face reconstruction to enable a label-efficient training strategy.In particular, we propose a Reinforcement Active Learning approach in conjunction with a clustering-based pooling strategy to select informative view-points of the subjects. Experimental results on 300W-LP and AFLW2000 datasets demonstrate that our proposed method is able to 1) efficiently select the most influencing view-points for labeling and outperforms several baseline AL techniques and 2) further improve the performance of a 3D Face Reconstruction network trained on the full dataset.",
author = "Hoda Mohaghegh and Hossein Rahmani and Hamid Laga and Farid Boussaid and Mohammed Bennamoun",
year = "2023",
month = jan,
day = "16",
language = "English",
booktitle = "IEEE International Conference on Robotics and Automation (ICRA)",

}

RIS

TY - GEN

T1 - Reinforced Learning for Label-Efficient 3D Face Reconstruction

AU - Mohaghegh, Hoda

AU - Rahmani, Hossein

AU - Laga, Hamid

AU - Boussaid, Farid

AU - Bennamoun, Mohammed

PY - 2023/1/16

Y1 - 2023/1/16

N2 - 3D face reconstruction plays a major role in many human-robot interaction systems, from automatic face authentication to human-computer interface-based entertainment. To improve robustness against occlusions and noise, 3D face reconstruction networks are often trained on a set of in-the-wild face images preferably captured along different viewpoints of the subject. However, collecting the required large amounts of 3D annotated face data is expensive and time-consuming. To address the high annotation cost and due to the importance of training on a useful set, we propose an Active Learning (AL) framework that actively selects the most informative and representative samples to be labeled. To the best of our knowledge, this paper is the first work on tackling active learning for 3D face reconstruction to enable a label-efficient training strategy.In particular, we propose a Reinforcement Active Learning approach in conjunction with a clustering-based pooling strategy to select informative view-points of the subjects. Experimental results on 300W-LP and AFLW2000 datasets demonstrate that our proposed method is able to 1) efficiently select the most influencing view-points for labeling and outperforms several baseline AL techniques and 2) further improve the performance of a 3D Face Reconstruction network trained on the full dataset.

AB - 3D face reconstruction plays a major role in many human-robot interaction systems, from automatic face authentication to human-computer interface-based entertainment. To improve robustness against occlusions and noise, 3D face reconstruction networks are often trained on a set of in-the-wild face images preferably captured along different viewpoints of the subject. However, collecting the required large amounts of 3D annotated face data is expensive and time-consuming. To address the high annotation cost and due to the importance of training on a useful set, we propose an Active Learning (AL) framework that actively selects the most informative and representative samples to be labeled. To the best of our knowledge, this paper is the first work on tackling active learning for 3D face reconstruction to enable a label-efficient training strategy.In particular, we propose a Reinforcement Active Learning approach in conjunction with a clustering-based pooling strategy to select informative view-points of the subjects. Experimental results on 300W-LP and AFLW2000 datasets demonstrate that our proposed method is able to 1) efficiently select the most influencing view-points for labeling and outperforms several baseline AL techniques and 2) further improve the performance of a 3D Face Reconstruction network trained on the full dataset.

M3 - Conference contribution/Paper

BT - IEEE International Conference on Robotics and Automation (ICRA)

ER -