AI Paper Atlas

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

ModuSeg: Decoupling Object Discovery and Semantic Retrieval for Training-Free Weakly Supervised Segmentation

Qingze He, Fagui Liu, Dengke Zhang, Qingmao Wei, Quan Tang

ECCV 2026 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Weakly supervised semantic segmentation · 0 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 SMFA (soft-masked feature aggregation) (ablation)ModuSeg baseline · VOC valmIoU (%): 84.3 -> 84.6Improved
add SBP (semantic boundary purification) (ablation)ModuSeg baseline · VOC valmIoU (%): 84.3 -> 85.2Improved
add both SBP and SMFA (ablation)ModuSeg baseline · VOC valmIoU (%): 84.3 -> 86.3Improved
add logits-based filtering to feature bank (CorrCLIP, VOC) (ablation)CorrCLIP · VOC trainmIoU (%): 68.7 -> 78.8Improved
add logits-based filtering to feature bank (CorrCLIP, COCO) (ablation)CorrCLIP · COCO trainmIoU (%): 42.5 -> 54.6Improved
add logits-based filtering to feature bank (Mask Adapter) (ablation)Mask Adapter · VOC trainmIoU (%): 76.6 -> 82.4Improved
add logits-based filtering to feature bank (Mask Adapter) (ablation)Mask Adapter · COCO trainmIoU (%): 55.1 -> 60.2Improved
use stronger backbone C-RADIOv4 SO400M instead of DINOv2 ViT-B (ablation)ModuSeg · VOC valmIoU (%): 82.9 -> 86.3Improved
use EntitySeg proposals instead of SAM 2 (ablation)ModuSeg · VOC valmIoU (%): 82.6 -> 86.3Improved
use EntitySeg proposals instead of SAM 2 (ablation)ModuSeg · COCO valmIoU (%): 54.8 -> 56.7Improved
apply ModuSeg to CorrCLIP versus direct zero-shotCorrCLIP · VOCmIoU (%): 74.8 -> 86.3Improved
apply ModuSeg to Mask Adapter versus direct zero-shotMask Adapter · VOCmIoU (%): 81.2 -> 86.6Improved

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