Advanced search
1 file | 1.15 MB Add to list

Self-supervised learning for robust object retrieval without human annotations

(2023) COMPUTERS & GRAPHICS-UK. 115. p.13-24
Author
Organization
Project
Abstract
This paper explores the potential of self-supervised learning as an alternative to supervised learning in the context of geometry-based 3D object retrieval. With the ongoing digitalization of many industries, an exponentially increasing number of 3D objects are processed by retrieval systems. In order to support new shapes, modern deep learning-based retrieval systems require retraining. The dominant paradigm for optimizing neural networks in this field is supervised classification training. Supervised learning requires time-consuming and expensive data annotation. Moreover, training neural networks for classification introduces a bias towards the classes in the training data, which is undesirable for retrieval systems encountering unseen object types in the wild. Through extensive experiments, we make a direct comparison between supervised and self-supervised learning on four datasets from three different domains (household, manufacturing and medical). For object classes seen during training, self-supervised and supervised learning are competitive. For unseen classes, self-supervised learning outperforms supervised learning in many cases. We conclude that self-supervised learning provides a powerful tool for circumventing labeling costs and providing more robust retrieval systems.
Keywords
Object retrieval, self-supervised learning, cross-domain

Downloads

  • 1-s2.0-S0097849323001231-main.pdf
    • full text (Published version)
    • |
    • open access
    • |
    • PDF
    • |
    • 1.15 MB

Citation

Please use this url to cite or link to this publication:

MLA
Van den Herrewegen, Jarne, et al. “Self-Supervised Learning for Robust Object Retrieval without Human Annotations.” COMPUTERS & GRAPHICS-UK, vol. 115, 2023, pp. 13–24, doi:10.1016/j.cag.2023.06.029.
APA
Van den Herrewegen, J., Tourwé, T., & wyffels, F. (2023). Self-supervised learning for robust object retrieval without human annotations. COMPUTERS & GRAPHICS-UK, 115, 13–24. https://doi.org/10.1016/j.cag.2023.06.029
Chicago author-date
Van den Herrewegen, Jarne, Tom Tourwé, and Francis wyffels. 2023. “Self-Supervised Learning for Robust Object Retrieval without Human Annotations.” COMPUTERS & GRAPHICS-UK 115: 13–24. https://doi.org/10.1016/j.cag.2023.06.029.
Chicago author-date (all authors)
Van den Herrewegen, Jarne, Tom Tourwé, and Francis wyffels. 2023. “Self-Supervised Learning for Robust Object Retrieval without Human Annotations.” COMPUTERS & GRAPHICS-UK 115: 13–24. doi:10.1016/j.cag.2023.06.029.
Vancouver
1.
Van den Herrewegen J, Tourwé T, wyffels F. Self-supervised learning for robust object retrieval without human annotations. COMPUTERS & GRAPHICS-UK. 2023;115:13–24.
IEEE
[1]
J. Van den Herrewegen, T. Tourwé, and F. wyffels, “Self-supervised learning for robust object retrieval without human annotations,” COMPUTERS & GRAPHICS-UK, vol. 115, pp. 13–24, 2023.
@article{01H9WHRHWB046HZNG8NSQ0XESZ,
  abstract     = {{This paper explores the potential of self-supervised learning as an alternative to supervised learning in the context of geometry-based 3D object retrieval. With the ongoing digitalization of many industries, an exponentially increasing number of 3D objects are processed by retrieval systems. In order to support new shapes, modern deep learning-based retrieval systems require retraining. The dominant paradigm for optimizing neural networks in this field is supervised classification training. Supervised learning requires time-consuming and expensive data annotation. Moreover, training neural networks for classification introduces a bias towards the classes in the training data, which is undesirable for retrieval systems encountering unseen object types in the wild. Through extensive experiments, we make a direct comparison between supervised and self-supervised learning on four datasets from three different domains (household, manufacturing and medical). For object classes seen during training, self-supervised and supervised learning are competitive. For unseen classes, self-supervised learning outperforms supervised learning in many cases. We conclude that self-supervised learning provides a powerful tool for circumventing labeling costs and providing more robust retrieval systems.}},
  author       = {{Van den Herrewegen, Jarne and Tourwé, Tom and wyffels, Francis}},
  issn         = {{0097-8493}},
  journal      = {{COMPUTERS & GRAPHICS-UK}},
  keywords     = {{Object retrieval,self-supervised learning,cross-domain}},
  language     = {{eng}},
  pages        = {{13--24}},
  title        = {{Self-supervised learning for robust object retrieval without human annotations}},
  url          = {{http://doi.org/10.1016/j.cag.2023.06.029}},
  volume       = {{115}},
  year         = {{2023}},
}

Altmetric
View in Altmetric
Web of Science
Times cited: