We introduce AnyUp, a method for feature upsampling that can be applied to any vision feature at any resolution, without encoder-specific training. Existing learning-based upsamplers for features like DINO or CLIP need to be re-trained for every feature extractor and thus do not generalize to different feature types at inference time. In this work, we propose an inference-time feature-agnostic upsampling architecture to alleviate this limitation and improve upsampling quality. In our experiments, AnyUp sets a new state of the art for upsampled features, generalizes to different feature types, and preserves feature semantics while being efficient and easy to apply to a wide range of downstream tasks.
@article{arxiv.2510.12764,
title = {AnyUp: Universal Feature Upsampling},
author = {Thomas Wimmer and Prune Truong and Marie-Julie Rakotosaona and Michael Oechsle and Federico Tombari and Bernt Schiele and Jan Eric Lenssen},
journal= {arXiv preprint arXiv:2510.12764},
year = {2026}
}
Comments
Accepted to ICLR 2026 (Oral). Project Website: https://wimmerth.github.io/anyup/