Dynamically Visual Disambiguation of Keyword-Based Image Search
Abstract
Due to the high cost of manual annotation, learning directly from the web has attracted broad attention. One issue that limits their performance is the problem of visual polysemy. To address this issue, we present an adaptive multi-model framework that resolves polysemy by visual disambiguation. Compared to existing methods, the primary advantage of our approach lies in that our approach can adapt to the dynamic changes in the search results. Our proposed framework consists of two major steps: we first discover and dynamically select the text queries according to the image search results, then we employ the proposed saliency-guided deep multi-instance learning network to remove outliers and learn classification models for visual disambiguation. Extensive experiments demonstrate the superiority of our proposed approach.
Cite
Text
Yao et al. "Dynamically Visual Disambiguation of Keyword-Based Image Search." International Joint Conference on Artificial Intelligence, 2019. doi:10.24963/IJCAI.2019/140Markdown
[Yao et al. "Dynamically Visual Disambiguation of Keyword-Based Image Search." International Joint Conference on Artificial Intelligence, 2019.](https://mlanthology.org/ijcai/2019/yao2019ijcai-dynamically/) doi:10.24963/IJCAI.2019/140BibTeX
@inproceedings{yao2019ijcai-dynamically,
title = {{Dynamically Visual Disambiguation of Keyword-Based Image Search}},
author = {Yao, Yazhou and Sun, Zeren and Shen, Fumin and Liu, Li and Wang, Limin and Zhu, Fan and Ding, Lizhong and Wu, Gangshan and Shao, Ling},
booktitle = {International Joint Conference on Artificial Intelligence},
year = {2019},
pages = {996-1002},
doi = {10.24963/IJCAI.2019/140},
url = {https://mlanthology.org/ijcai/2019/yao2019ijcai-dynamically/}
}