This website requires JavaScript.

Deep Multimodal Fusion for Generalizable Person Re-identification

Suncheng XiangHao ChenJingsheng Gao ...+4 Yuzhuo Fu
Nov 2022
摘要
Person re-identification plays a significant role in realistic scenarios dueto its various applications in public security and video surveillance.Recently, leveraging the supervised or semi-unsupervised learning paradigms,which benefits from the large-scale datasets and strong computing performance,has achieved a competitive performance on a specific target domain. However,when Re-ID models are directly deployed in a new domain without target samples,they always suffer from considerable performance degradation and poor domaingeneralization. To address this challenge, in this paper, we propose DMF, aDeep Multimodal Fusion network for the general scenarios on personre-identification task, where rich semantic knowledge is introduced to assistin feature representation learning during the pre-training stage. On top of it,a multimodal fusion strategy is introduced to translate the data of differentmodalities into the same feature space, which can significantly boostgeneralization capability of Re-ID model. In the fine-tuning stage, a realisticdataset is adopted to fine-tine the pre-trained model for distributionalignment with real-world. Comprehensive experiments on benchmarks demonstratethat our proposed method can significantly outperform previous domaingeneralization or meta-learning methods. Our source code will also be publiclyavailable at https://github.com/JeremyXSC/DMF.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答