AffectSeek is an agentic framework that localizes affective moments, classifies emotions, and generates rationales in long videos under vague user queries, backed by the new VQAU-Bench benchmark.
Mer 2023: Multi-label learning, modality robustness, and semi-supervised learning,
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
-
AffectSeek: Agentic Affective Understanding in Long Videos under Vague User Queries
AffectSeek is an agentic framework that localizes affective moments, classifies emotions, and generates rationales in long videos under vague user queries, backed by the new VQAU-Bench benchmark.