Pedestrian attribute recognition based on multiple time steps attention

Zhong Ji, Zhenfei Hu,Erlu He,Jungong Han,Yanwei Pang

Pattern Recognition Letters（2020）

引用 20|浏览32

暂无评分

摘要

•A MTA-Net framework is designed for Pedestrian Attribute Recognition (PAR).•A novel attention mechanism for PAR is proposed to exploit more contextual knowledge.•A focal balance loss is developed to alleviate the attribute imbalance issue.

查看译文

关键词

Pedestrian attribute,Attention network,Intelligent surveillance

AI 理解论文

溯源树

样例

生成溯源树，研究论文发展脉络

Chat Paper

正在生成论文摘要