Hand-object masked training and an HOI-dynamics-aware decoder yield more balanced cue-specific action recognition on a new inpainted DEHOI testbed and transfer to object-state and robot-manipulation tasks.
In: European Conference on Computer Vision
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
Hand-object masked training and an HOI-dynamics-aware decoder yield more balanced cue-specific action recognition on a new inpainted DEHOI testbed and transfer to object-state and robot-manipulation tasks.