MCFL: multi-label contrastive focal loss for deep imbalanced pedestrian attribute recognition
文献类型：期刊论文
| 作者 | Chen, Lin1 ; Song, Jingkuan2; Zhang, Xuerui1; Shang, Mingsheng1
|
| 刊名 | NEURAL COMPUTING & APPLICATIONS
 |
| 出版日期 | 2022-06-04 |
| 页码 | 15 |
| 关键词 | Pedestrian attribute recognition Multi-label contrastive loss Deep convolutional neural network Multi-label learning Imbalanced learning |
| ISSN号 | 0941-0643 |
| DOI | 10.1007/s00521-022-07300-7 |
| 通讯作者 | Shang, Mingsheng(msshang@cigit.ac.cn) |
| 英文摘要 | Pedestrian Attribute Recognition (PAR) can provide valuable clues for several innovative surveillance applications. It is also a difficult task because inference of the multiple attributes at a far distance is challenging in real complex scenarios. Most existing methods improve the PAR with visual attention mechanisms or body-part detection modules, which increase the complexity of networks and require manual annotations of the human body. Also, uneven data distribution, leading to a decline in recall values, is still underestimated. This paper presents a novel multi-label optimization algorithm to mitigate these issues, named Multi-label Contrastive Focal Loss (MCFL). Specifically, we first propose a multi-label focal loss to emphasize the error-prone and minority attributes with a separated re-weighting scheme. And then, we introduce a multi-label contrastive learning strategy based on the multi-label divergences to help the deep network to distinguish the hard fine-grained attributes. We conduct extensive experiments on seven PAR benchmarks, and results indicate that the proposed MCFL with the native ResNet-50 backbone surpasses the state-of-the-art comparison methods in mean accuracy and recall. |
| 资助项目 | National Nature Science Foundation of China[61902370] ; Chongqing Research Program of Technology Innovation and Application[cstc2019jscx-zdztzxX0019] ; Chongqing Municipal Eduction Commission[HZ2021008] ; Chongqing Municipal Eduction Commission[HZ2021017] |
| WOS研究方向 | Computer Science |
| 语种 | 英语 |
| WOS记录号 | WOS:000805925400004 |
| 出版者 | SPRINGER LONDON LTD |
| 源URL | [http://119.78.100.138/handle/2HOD01W0/16311]  |
| 专题 | 中国科学院重庆绿色智能技术研究院 |
| 通讯作者 | Shang, Mingsheng |
| 作者单位 | 1.Chinese Acad Sci, Chongqing Inst Green & Intelligent Technol, Chongqing 400714, Peoples R China 2.Univ Elect Sci & Technol China, Chengdu 611731, Sichuan, Peoples R China |
| 推荐引用方式 GB/T 7714 | Chen, Lin,Song, Jingkuan,Zhang, Xuerui,et al. MCFL: multi-label contrastive focal loss for deep imbalanced pedestrian attribute recognition[J]. NEURAL COMPUTING & APPLICATIONS,2022:15. |
| APA | Chen, Lin,Song, Jingkuan,Zhang, Xuerui,&Shang, Mingsheng.(2022).MCFL: multi-label contrastive focal loss for deep imbalanced pedestrian attribute recognition.NEURAL COMPUTING & APPLICATIONS,15. |
| MLA | Chen, Lin,et al."MCFL: multi-label contrastive focal loss for deep imbalanced pedestrian attribute recognition".NEURAL COMPUTING & APPLICATIONS (2022):15. |
入库方式： OAI收割
来源：重庆绿色智能技术研究院
浏览0
下载0
收藏0
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。