.png){kind=icon}
.png){kind=icon}
Return
Equidistant deep embedding-based multi-label group activity recognition with dependency-constrained training
.png){kind=icon}
PRE
AI
DOI：10.1016/j.asoc.2025.113721.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Novel multi-label GAR approach with similar single-label learning. • EDE network employs self-supervision for equidistant embeddings. • DLC strategy optimizes neural networks independent of classification layers. • Auxiliary TransOvR module enhances training for inter-activity dependency. • Better model performance with less parameters.
Keywords:
multi-label GAR
self-supervision
equidistant embeddings
DLC strategy
TransOvR module
Journal
IF:
6.6
Papers:
1.4W
Citations:
4.8W