Back to Search
Start Over
MCFL: multi-label contrastive focal loss for deep imbalanced pedestrian attribute recognition.
- Source :
-
Neural Computing & Applications . Oct2022, Vol. 34 Issue 19, p16701-16715. 15p. - Publication Year :
- 2022
-
Abstract
- 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. [ABSTRACT FROM AUTHOR]
Details
- Language :
- English
- ISSN :
- 09410643
- Volume :
- 34
- Issue :
- 19
- Database :
- Academic Search Index
- Journal :
- Neural Computing & Applications
- Publication Type :
- Academic Journal
- Accession number :
- 159263318
- Full Text :
- https://doi.org/10.1007/s00521-022-07300-7