CMDCL debiases text embeddings by back-door adjustment and deconfounds video features by front-door adjustment, achieving state-of-the-art long-term action recognition on three benchmarks.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Cross-Modal Dual-Causal Learning for Long-Term Action Recognition
CMDCL debiases text embeddings by back-door adjustment and deconfounds video features by front-door adjustment, achieving state-of-the-art long-term action recognition on three benchmarks.