EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Siamese comparative transformer-based network for unsupervised landmark detection

Can Zhao iD, Tao Wu, Jianlin Zhang, Zhiyong Xu, Meihui Li, Dongxu Liu

PLOS ONE · 2024

DOI: 10.1371/journal.pone.0313518 ↗

Kurz gefasst

Fachartikel in der Zeitschrift PLOS ONE, 2024. Verfasst von Can Zhao und 5 weiteren Autor:innen. Beteiligte Einrichtung: National Key Laboratory of Optical Field Manipulation Science and Technology, Chinese Academy of Sciences, Chengdu, China. Hauptthemen laut MEDLINE: Algorithms, Image Processing, Computer-Assisted. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China under Grant, Postdoctoral Fellowship Program of CPSF.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Landmark detection is a common task that benefits downstream computer vision tasks. Current landmark detection algorithms often train a sophisticated image pose encoder by reconstructing the source image to identify landmarks. Although a well-trained encoder can effectively capture landmark information through image reconstruction, it overlooks the semantic relationships between landmarks. This contradicts the goal of achieving semantic representations in landmark detection tasks. To address these challenges, we introduce a novel Siamese comparative transformer-based network that strengthens the semantic connections among detected landmarks. Specifically, the connection between landmarks with the same semantics has been enhanced by employing a Siamese contrastive regularizer. In addition, we integrate a lightweight direction-guided Transformer into the image pose encoder to perceive global feature relationships, thereby improving the representation and encoding of landmarks. Experiments on the CelebA, AFLW, and Cat Heads benchmarks demonstrate that our proposed method achieves competitive performance compared to existing unsupervised methods and even supervised methods.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

AlgorithmsImage Processing, Computer-AssistedAnimalsCatsHumansSemanticsNeural Networks, ComputerHead

Autor:innen und Institutionen

  • National Key Laboratory of Optical Field Manipulation Science and Technology, Chinese Academy of Sciences, Chengdu, China.
  • Key Laboratory of Optical Engineering, Chinese Academy of Sciences, Chengdu, China.
  • Institute of Optics and Electronics, Chinese Academy of Sciences, Chengdu, China.
  • University of the Chinese Academy of Science, Beijing, China.

ORCID-Profile: Can Zhao

Förderung

  • National Natural Science Foundation of China under Grant
  • Postdoctoral Fellowship Program of CPSF

Zitieren

APA

Zhao, C., Wu, T., Zhang, J., Xu, Z., Li, M., & Liu, D. (2024). Siamese comparative transformer-based network for unsupervised landmark detection. PLOS ONE. https://doi.org/10.1371/journal.pone.0313518

Vancouver

Zhao C, Wu T, Zhang J, Xu Z, Li M, Liu D. Siamese comparative transformer-based network for unsupervised landmark detection. PLOS ONE. 2024. doi:10.1371/journal.pone.0313518

BibTeX
@article{Zhao2024,
  author = {Zhao, Can and Wu, Tao and Zhang, Jianlin and Xu, Zhiyong and Li, Meihui and Liu, Dongxu},
  title = {Siamese comparative transformer-based network for unsupervised landmark detection},
  journal = {PLOS ONE},
  year = {2024},
  doi = {10.1371/journal.pone.0313518}
}
Kostenlos starten →