Visual-Semantic Decomposition and Partial Alignment for Document-based Zero-Shot Learning
Abstract
Recent work shows that documents from encyclopedias serve as helpful auxiliary information for zero-shot learning. Existing methods align the entire semantics of a document with corresponding images to transfer knowledge. However, they disregard that semantic information is not equivalent between them, resulting in a suboptimal alignment. In this work, we propose a novel network to extract multi-view semantic concepts from documents and images and align the matching rather than entire concepts. Specifically, we propose a semantic decomposition module to generate multi-view semantic embeddings from visual and textual sides, providing the basic concepts for partial alignment. To alleviate the issue of information redundancy among embeddings, we propose the local-to-semantic variance loss to capture distinct local details and multiple semantic diversity loss to enforce orthogonality among embeddings. Subsequently, two losses are introduced to partially align visual-semantic embedding pairs according to their semantic relevance at the view and word-to-patch levels. Consequently, we consistently outperform state-of-the-art methods under two document sources in three standard benchmarks for document-based zero-shot learning. Qualitatively, we show that our model learns the interpretable partial association.
Cite
@article{arxiv.2407.15613,
title = {Visual-Semantic Decomposition and Partial Alignment for Document-based Zero-Shot Learning},
author = {Xiangyan Qu and Jing Yu and Keke Gai and Jiamin Zhuang and Yuanmin Tang and Gang Xiong and Gaopeng Gou and Qi Wu},
journal= {arXiv preprint arXiv:2407.15613},
year = {2024}
}
Comments
Accepted to ACM International Conference on Multimedia (MM) 2024