VGGT-Ω improves feed-forward reconstruction accuracy and efficiency by architectural simplifications, register-based attention, and training on much larger supervised and unlabeled video data.
Zhai, and Shenlong Wang
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
-
VGGT-$\Omega$
VGGT-Ω improves feed-forward reconstruction accuracy and efficiency by architectural simplifications, register-based attention, and training on much larger supervised and unlabeled video data.