Upsample Anything: A Simple and Hard to Beat Baseline for Feature Upsampling

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Seo, Minseok, Hamilton, Mark, Kim, Changick
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912725984083968
author Seo, Minseok
Hamilton, Mark
Kim, Changick
author_facet Seo, Minseok
Hamilton, Mark
Kim, Changick
contents We present \textbf{Upsample Anything}, a lightweight test-time optimization (TTO) framework that restores low-resolution features to high-resolution, pixel-wise outputs without any training. Although Vision Foundation Models demonstrate strong generalization across diverse downstream tasks, their representations are typically downsampled by 14x/16x (e.g., ViT), which limits their direct use in pixel-level applications. Existing feature upsampling approaches depend on dataset-specific retraining or heavy implicit optimization, restricting scalability and generalization. Upsample Anything addresses these issues through a simple per-image optimization that learns an anisotropic Gaussian kernel combining spatial and range cues, effectively bridging Gaussian Splatting and Joint Bilateral Upsampling. The learned kernel acts as a universal, edge-aware operator that transfers seamlessly across architectures and modalities, enabling precise high-resolution reconstruction of features, depth, or probability maps. It runs in only $\approx0.419 \text{s}$ per 224x224 image and achieves state-of-the-art performance on semantic segmentation, depth estimation, and both depth and probability map upsampling. \textbf{Project page:} \href{https://seominseok0429.github.io/Upsample-Anything/}{https://seominseok0429.github.io/Upsample-Anything/}
format Preprint
id arxiv_https___arxiv_org_abs_2511_16301
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Upsample Anything: A Simple and Hard to Beat Baseline for Feature Upsampling
Seo, Minseok
Hamilton, Mark
Kim, Changick
Computer Vision and Pattern Recognition
We present \textbf{Upsample Anything}, a lightweight test-time optimization (TTO) framework that restores low-resolution features to high-resolution, pixel-wise outputs without any training. Although Vision Foundation Models demonstrate strong generalization across diverse downstream tasks, their representations are typically downsampled by 14x/16x (e.g., ViT), which limits their direct use in pixel-level applications. Existing feature upsampling approaches depend on dataset-specific retraining or heavy implicit optimization, restricting scalability and generalization. Upsample Anything addresses these issues through a simple per-image optimization that learns an anisotropic Gaussian kernel combining spatial and range cues, effectively bridging Gaussian Splatting and Joint Bilateral Upsampling. The learned kernel acts as a universal, edge-aware operator that transfers seamlessly across architectures and modalities, enabling precise high-resolution reconstruction of features, depth, or probability maps. It runs in only $\approx0.419 \text{s}$ per 224x224 image and achieves state-of-the-art performance on semantic segmentation, depth estimation, and both depth and probability map upsampling. \textbf{Project page:} \href{https://seominseok0429.github.io/Upsample-Anything/}{https://seominseok0429.github.io/Upsample-Anything/}
title Upsample Anything: A Simple and Hard to Beat Baseline for Feature Upsampling
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.16301