AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Progressive Proxy Anchor Propagation for Unsupervised Semantic Segmentation

Hyun Seok Seong, WonJun Moon, SuBeen Lee, Jae-Pil Heo

ECCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Unsupervised object and part discovery · 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
add trustworthy positive set via proxy anchor propagation (TPS) (ablation)PPAP contrastive baseline · COCO-stuff ViT-S/16unsupervised mIoU: 22.9 -> 25.0Improved
add ambiguity-excluded negative set (ANS) (ablation)PPAP contrastive baseline · COCO-stuff ViT-S/16unsupervised mIoU: 22.9 -> 25.1Improved
combine TPS and ANS (ablation)PPAP · COCO-stuff ViT-S/16unsupervised accuracy: 49.1 -> 62.9Improved
apply CRF post-processing (ablation)PPAP · COCO-stuff ViT-S/16unsupervised accuracy: 59.9 -> 62.9Improved
replace HP with PPAP, both without CRFCOCO-stuff ViT-S/16 · COCO-stuffunsupervised accuracy: 54.5 -> 59.9Improved
use PPAP instead of HP on larger-patch ViT-S/16 backboneunsupervised segmentation · COCO-stuffunsupervised accuracy: 54.5 -> 62.9Improved
use PPAP instead of HP on Potsdam-3 (only 3 classes)ViT-B/8 · Potsdam-3unsupervised mIoU: 69.7 -> 71.0Improved
use PPAP instead of HP on Cityscapes ViT-S/8 linear probeViT-S/8 · Cityscapeslinear accuracy: 91.2 -> 90.8Hurt
use PPAP instead of HP on ImageNet-SViT-S/16 · ImageNet-S valmIoU: 8.6 -> 25.7Improved
apply PPAP on iBoT ViT-B/16 backbone instead of HPiBoT · COCO-stuffunsupervised mIoU: 22.4 -> 27.6Improved
use PPAP instead of HP on few-shot (rare) classesCOCO-stuff · COCO-stuff class frequencymIoU (Few): 28.9 -> 32.9Improved
relocate proxy anchor progressively (PPAP) instead of k-means grouping for positivespositive set selection · COCO-stuff ViT-S/16positive precision: 22% -> 72%Improved

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