A coordinate-based cross-attention transformer, trained with mask-refined and self-distilled pseudo-groundtruth, upsamples VFM features to full resolution and improves downstream task performance.
Lanczos filtering in one and two dimen- sions
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
-
LoftUp: Learning a Coordinate-Based Feature Upsampler for Vision Foundation Models
A coordinate-based cross-attention transformer, trained with mask-refined and self-distilled pseudo-groundtruth, upsamples VFM features to full resolution and improves downstream task performance.