AI Paper Atlas

Research map / Image generation / Fast diffusion sampling

SegRefiner: Towards Model-Agnostic Segmentation Refinement with Discrete Diffusion Process

Mengyu Wang, Henghui Ding, Jun Hao Liew, Jiajun Liu, Yao Zhao, Yunchao Wei

NeurIPS 2023 · Image generation › Fast diffusion sampling · approach: Diffusion models as perception tools · 6 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
add discrete diffusion sampling to iterative refinement (vs none) (ablation)SegRefiner · BIGmBA: 75.03 -> 77.64Improved
add discrete diffusion sampling to iterative refinement (vs none) (ablation)SegRefiner · BIGIoU: 94.10 -> 94.85Improved
iterate with previous mask as input but no diffusion sampling (ablation)SegRefiner · BIGmBA: 75.03 -> 75.10No clear effect
add local refinement to global refinement (ablation)SegRefiner · BIGmBA: 60.58 -> 77.64Improved
add global refinement to local refinement (ablation)SegRefiner · BIGIoU: 89.73 -> 94.85Improved
add local refinement to global refinement (ablation)SegRefiner · BIGIoU: 92.43 -> 94.85Improved
input size 128x128 to 256x256 (ablation)SegRefiner · BIGIoU: 93.10 -> 94.85Improved
input size 256x256 to 512x512 (ablation)SegRefiner · BIGmBA: 77.64 -> 77.56No clear effect
increase iteration steps from 1 to 6 (ablation)SegRefiner · COCO val (LVIS), Mask R-CNN R50Boundary AP: 29.9 -> 32.6Improved
increase iteration steps from 5 to 6 (ablation)SegRefiner · COCO val (LVIS), Mask R-CNN R50Mask AP: 41.9 -> 41.9No clear effect
refine Mask R-CNN R50 with SegRefiner instead of BPRMask R-CNN R50 · COCO val (LVIS)Boundary AP: 30.4 -> 32.6Improved
apply SegRefiner to SOLOSOLO · COCO val (LVIS)Boundary AP: 24.7 -> 31.3Improved

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