AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

S2-UniSeg: Fast Universal Agglomerative Pooling for Scalable Segment Anything Without Supervision

Huihui Xu, Jin Ye, Hongqiu Wang, Changkai Ji, Jiashi Lin, Ming Hu, Ziyan Huang, Ying Chen, Chenglong Ma, Tianbin Li, Lihao Liu, Junjun He and 1 more

AAAI 2026 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Weakly supervised semantic segmentation · 1 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 spatial similarity (omega_s=0.4) to the identify step (ablation)UniAP · UVO valAP: 12.4 -> 16.2Improved
use much lower thresholds (0.5-0.1) (ablation)UniAP · UVO valAP: 16.2 -> 15.5Hurt
use finer thresholds (0.9-0.1) (ablation)UniAP · UVO valAP: 16.2 -> 16.7Mixed
use 4 local crops instead of 2 (ablation)S2-UniSeg · UVO valAP: 16.2 -> 15.9No clear effect
use 6 local crops instead of 2 (ablation)S2-UniSeg · UVO valAR100: 32.1 -> 32.3No clear effect
increase QuerySD projection dimension K from 512 to 1024 (ablation)S2-UniSeg · UVO valAP: 16.2 -> 16.9Improved
reduce K from 512 to 128 (ablation)S2-UniSeg · UVO valAP: 16.2 -> 13.6Hurt
scale pretraining data from 0.4M to 2.0M SA-1B imagesS2-UniSeg · COCO class-agnostic instance segAP: 33.6 -> 36.4Improved
scale pretraining data from 0.4M to 2.0M SA-1B imagesS2-UniSeg · COCO class-agnostic instance segAR: 43.9 -> 46.3Improved
replace multi-stage offline pseudo-masks with fast UniAP and single-stage pretrainingS2-UniSeg vs U2Seg · COCO class-agnostic instance segAP: 7.3 -> 14.2Improved
use UniAP instead of MaskCut for pseudo-maskspseudo-mask generation · COCO val2017time per image (s): 4.72 -> 0.045Improved
use UniAP instead of MaskCut for pseudo-maskspseudo-mask generation · COCO val2017AP mask: 6.7 -> 6.2Hurt

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