Interactive segmentation reveals that feature upsampler choice strongly affects dense prediction quality in frozen DINOv2, with LoftUp giving the best results.
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
-
Benchmarking Feature Upsampling Methods for Vision Foundation Models using Interactive Segmentation
Interactive segmentation reveals that feature upsampler choice strongly affects dense prediction quality in frozen DINOv2, with LoftUp giving the best results.