Entanglement Loss for Context-Based Still Image Action Recognition
文献类型:会议论文
作者 | Xin M(辛淼)1; Shuhang Wang2; Jian Cheng1 |
出版日期 | 2019 |
会议日期 | July 8, 2019 - July 12, 2019 |
会议地点 | Shanghai, China |
关键词 | Still image action recognition attribute entanglement feature learning loss function |
DOI | 10.1109/ICME.2019.00183 |
页码 | 1042-1047 |
英文摘要 | We observed an attribute entanglement phenomenon: samples with similar attributes but from different classes can easily result in recognition errors. This problem is an important cause that results in recognition errors. To address this problem, we propose a new loss function, namely the entanglement loss. It penalizes the compactness between the misclassified entangled samples and their misclassified class centers, such that the features of entangled samples are apart from the misclassified classes. The proposed loss function can effectively enhance the discriminative power of the deeply learned features, thus recognition performance can be significantly improved. Experimental results show that our method outperforms the previous state-of-the-art methods on PASCAL VOC 2012 Action and ASLAN datasets. |
语种 | 英语 |
WOS记录号 | WOS:000501820600175 |
源URL | [http://ir.ia.ac.cn/handle/173211/51505] |
专题 | 复杂系统认知与决策实验室 |
通讯作者 | Xin M(辛淼) |
作者单位 | 1.Institute of Automation, Chinese Academy of Sciences 2.Harvard University |
推荐引用方式 GB/T 7714 | Xin M,Shuhang Wang,Jian Cheng. Entanglement Loss for Context-Based Still Image Action Recognition[C]. 见:. Shanghai, China. July 8, 2019 - July 12, 2019. |
入库方式: OAI收割
来源:自动化研究所
其他版本
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。