Researcher, Thales SIX
1 paper at NeurIPS 2025
We propose a novel architecture and training objective specifically designed to upsample features from foundation vision encoders at any resolution.
