AI Paper Atlas

Research map / Detection and segmentation / CNN features

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

Minseok Seo, Mark Hamilton, Changick Kim

CVPR 2026 · Detection and segmentation › CNN features · approach: Context for hard detection conditions · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use per-image anisotropic Gaussian TTO upsampling instead of bilinearDINOv2-S linear probe · COCOmIoU: 60.43 -> 61.41Improved
use per-image anisotropic Gaussian TTO upsampling instead of bilinearDINOv2-S linear probe · ADE20kmIoU: 41.48 -> 42.95Improved
use per-image anisotropic Gaussian TTO upsampling instead of bilinearDINOv2-S linear probe · PASCAL-VOCmIoU: 81.27 -> 82.22Improved
upsample predicted probability maps instead of featuresUpsample Anything · PASCAL-VOCmIoU: 82.22 -> 84.57Improved
upsample predicted probability maps instead of featuresUpsample Anything · CityscapesmIoU: 57.92 -> 56.36Hurt
feature upsampling methods versus bilinear under strong backbone and long head trainingall upsamplers · CityscapesmIoU: 57.90 -> 57.92No clear effect
use Upsample Anything instead of bilinear for depthDINOv2 frozen, DPT-style head · NYUv2 depth estimationRMSE: 0.545 -> 0.498Improved
use LoftUp upsampling instead of bilinearDINOv2 DPT head · NYUv2 depth estimationRMSE: 0.545 -> 0.796Hurt
use Upsample Anything instead of bilinear for depth map upsampling32x32 to 512x512 depth · MiddleburyRMSE: 0.231 -> 0.209Improved
use Upsample Anything instead of bilinear for depth map upsampling32x32 to 512x512 depth · NYUv2 depth map upsamplingRMSE: 0.167 -> 0.214Hurt
add per-pixel anisotropic Gaussian kernels with TTO (GSJBU) over fixed-kernel JBU (ablation)JBU · PASCAL-VOCmIoU: 81.65 -> 82.22Improved
add spatial-range constraint (GSJBU) over Guided Linear Upsample (ablation)Guided Linear Upsample · NYUv2 depth estimationRMSE: 0.598 -> 0.498Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy