Separate memory banks for shallow pixel details and deep semantic cues, merged by asymmetric cross-attention modules, improve unsupervised video object segmentation on DAVIS-16, FBMS, and YouTube-Objects.
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
CONDITIONAL 1representative citing papers
citing papers explorer
-
Shallow Features Matter: Hierarchical Memory with Heterogeneous Interaction for Unsupervised Video Object Segmentation
Separate memory banks for shallow pixel details and deep semantic cues, merged by asymmetric cross-attention modules, improve unsupervised video object segmentation on DAVIS-16, FBMS, and YouTube-Objects.