Transductive Zero-Shot Learning via Visual Center Adaptation

AAAI 2019 pp. 10059-10060

doi:10.1609/AAAI.V33I01.330110059 /aaai/2019/wan2019aaai-transductive/

Abstract

In this paper, we propose a Visual Center Adaptation Method (VCAM) to address the domain shift problem in zero-shot learning. For the seen classes in the training data, VCAM builds an embedding space by learning the mapping from semantic space to some visual centers. While for unseen classes in the test data, the construction of embedding space is constrained by a symmetric Chamfer-distance term, aiming to adapt the distribution of the synthetic visual centers to that of the real cluster centers. Therefore the learned embedding space can generalize the unseen classes well. Experiments on two widely used datasets demonstrate that our model significantly outperforms state-of-the-art methods.

AAAI Semantic Scholar

Cite

Text

Wan et al. "Transductive Zero-Shot Learning via Visual Center Adaptation." AAAI Conference on Artificial Intelligence, 2019. doi:10.1609/AAAI.V33I01.330110059

Markdown

[Wan et al. "Transductive Zero-Shot Learning via Visual Center Adaptation." AAAI Conference on Artificial Intelligence, 2019.](https://mlanthology.org/aaai/2019/wan2019aaai-transductive/) doi:10.1609/AAAI.V33I01.330110059

BibTeX

@inproceedings{wan2019aaai-transductive,
  title     = {{Transductive Zero-Shot Learning via Visual Center Adaptation}},
  author    = {Wan, Ziyu and Li, Yan and Yang, Min and Zhang, Junge},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2019},
  pages     = {10059-10060},
  doi       = {10.1609/AAAI.V33I01.330110059},
  url       = {https://mlanthology.org/aaai/2019/wan2019aaai-transductive/}
}