AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

Warp-Refine Propagation: Semi-Supervised Auto-Labeling via Cycle-Consistency

Aditya Ganeshan, Alexis Vallet, Yasunori Kudo, Shin-ichi Maeda, Tommi Kerola, Rares Ambrus, Dennis Park, Adrien Gaidon

ICCV 2021 · Label-efficient and robust learning › Semi-supervised learning · approach: Consistency training for semi-supervised segmentation · 15 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 learned refinement network to warp-inpaint labels (ablation)warp-inpaint · NYU-V2mIoU: 51.28 -> 52.12Improved
add learned refinement network to warp-inpaint labels (ablation)warp-inpaint · ApolloScapemIoU: 72.97 -> 73.74Improved
add semantic inpainting to motion-only labels (ablation)motion-only · NYU-V2mIoU: 50.38 -> 51.28Improved
train with warp-refine auto-labels instead of noneMSA-HRNet-OCR · NYU-V2mIoU: 50.58 -> 52.12Improved
train with warp-refine auto-labels instead of noneMSA-HRNet-OCR · ApolloScapemIoU: 72.63 -> 73.74Improved
train with warp-refine auto-labels instead of noneMSA-HRNet-OCR · Cityscapes valmIoU: 83.35 -> 84.07Improved
train with motion-only auto-labelsMSA-HRNet-OCR · Cityscapes valmIoU: 83.35 -> 83.01Hurt
train with motion-only auto-labelsMSA-HRNet-OCR · NYU-V2mIoU: 50.58 -> 50.38Hurt
train with semantic-only auto-labelsMSA-HRNet-OCR · NYU-V2mIoU: 50.58 -> 50.93Improved
use warp-refine instead of semantic-only on a high-performing baselineOCR Net ResNeXt-50 · Cityscapes val (no extra data)mIoU: 80.89 -> 80.80No clear effect
use warp-refine instead of semantic-onlyDeepLab V3 ResNeXt-50 · Cityscapes val (no extra data)mIoU: 80.45 -> 80.68Improved
use ground-truth labels instead of warp-refine propagated labels (oracle)MSA-HRNet-OCR · ApolloScapemIoU: 73.74 -> 75.36Improved

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