AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Deepstrip: High-Resolution Boundary Refinement

Peng Zhou, Brian Price, Scott Cohen, Gregg Wilensky, Larry S. Davis

CVPR 2020 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Cross-domain few-shot with transformers · 22 citations (OpenAlex)

Check 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
Use strip-based boundary prediction with the full selection and loss frameworkDeepStrip · DAVIS 2016 and PixaHR 8×, 16×, 32×Boundary F score: DAVIS 2016: 0.320 -> 0.423 (F (0 pix)); PixaHR 8×: 0.334 -> 0.416 (F (1 pix)); PixaHR 16×: 0.303 -> 0.396 (F (1 pix)); PImproved
Add Dice loss (ablation)U-Net strip · DAVIS 2016 and PixaHR 16×Boundary F score: 0.303 -> 0.323; 0.303 -> 0.320Improved
Add a selection layer to choose target boundaries (ablation)U-Net strip with Dice loss · DAVIS 2016 and PixaHR 16×Boundary F score: 0.323 -> 0.372; 0.320 -> 0.328Improved
Add boundary distance loss (ablation)U-Net strip with Dice loss and selection · DAVIS 2016 and PixaHR 16×Boundary F score: 0.372 -> 0.390; 0.328 -> 0.342Improved
Add C0 continuity regularization (ablation)DeepStrip without matching loss · DAVIS 2016 and PixaHR 16×Boundary F score: 0.390 -> 0.405; 0.342 -> 0.365Improved
Add matching loss (ablation)DeepStrip · DAVIS 2016 and PixaHR 16×Boundary F score: 0.405 -> 0.423; 0.365 -> 0.396Improved
Adapt strip height for one contour segment (ablation)DeepStrip · PixaHR 32×F (1 pix): 0.330 -> 0.353Improved
Adapt strip height independently for two segments (ablation)DeepStrip · PixaHR 32×F (1 pix): 0.353 -> 0.365Improved
Use image gradient alone for strip reconstruction instead of learned boundary prediction (ablation)DeepStrip reconstruction · DAVIS 2016 and PixaHR 16×Boundary F score: 0.423 -> 0.165; 0.396 -> 0.295Hurt
Vary strip width from factor 1 to 2, with factor 1.5 as the selected trade-off (ablation)DeepStrip · DAVIS 2016 and PixaHR 16×Boundary F score: DAVIS 2016: 0.414, 0.416, 0.423; PixaHR 16×: 0.392, 0.415, 0.396Mixed

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