The EAR method improves uni-modal event perception in audio-visual video parsing via targeted enhancements to pseudo-label generation and feature modeling, outperforming prior state-of-the-art approaches.
AMDF" denotes the Asymmetric Audio/Visual- Driven Fusion module, and
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
EAR: Enhancing Uni-Modal Representations for Weakly Supervised Audio-Visual Video Parsing
The EAR method improves uni-modal event perception in audio-visual video parsing via targeted enhancements to pseudo-label generation and feature modeling, outperforming prior state-of-the-art approaches.