AI Paper Atlas

Research map / Detection and segmentation / CNN features

Learning to Upsample by Learning to Sample

Wenze Liu, Hao Lu, Hongtao Fu, Zhiguo Cao

ICCV 2023 · Detection and segmentation › CNN features · approach: Joint detection and instance segmentation · 947 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 bilinear instead of nearest initial sampling positions (ablation)DySample · ADE20K (SegFormer-B1)mIoU: 41.9 -> 42.1Improved
multiply offset by static scope factor 0.25 (ablation)DySample · ADE20K (SegFormer-B1)mIoU: 42.1 -> 42.4Improved
strictly constrain offset scope with tanh instead of soft factorDySample · detection / segmentationAP / mIoUHurt
split features into 4 groups with separate offsets (ablation)DySample · ADE20K (SegFormer-B1)mIoU: 42.4 -> 43.2Improved
add dynamic scope factor (ablation)DySample · ADE20K (SegFormer-B1)mIoU: 43.2 -> 43.3Improved
use pixel shuffle + linear (PL) offset generation instead of linear + pixel shuffle (LP)DySample · SegFormer / MaskFormer vs other modelsmIoUMixed
replace bilinear upsampling with DySample-S+ in SegFormer-B1SegFormer-B1 · ADE20KmIoU: 41.68 -> 43.58Improved
replace nearest with DySample-S+ in MaskFormer Swin-BMaskFormer Swin-B · ADE20KmIoU: 52.70 -> 53.91Improved
replace nearest with DySample+ in Faster R-CNN R50Faster R-CNN R50 · MS COCOAP: 37.5 -> 38.7Improved
replace nearest with DySample+ in Panoptic FPN R50Panoptic FPN R50 · MS COCOPQ: 40.2 -> 41.5Improved
use DySample (no guidance) instead of guided upsamplers FADE/SAPA for boundary qualitySegFormer-B1 · ADE20KbIoU: 31.68 -> 29.93Hurt
replace bilinear with DySample+ in DepthFormerDepthFormer Swin-T · NYU Depth V2delta < 1.25: 0.873 -> 0.878Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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