AI Paper Atlas

Research map / Detection and segmentation / CNN features

Pixels, Regions, and Objects: Multiple Enhancement for Salient Object Detection

Yi Wang, Ruili Wang, Xin Fan, Tianzhu Wang, Xiangjian He

CVPR 2023 · Detection and segmentation › CNN features · approach: RGB-D salient object detection · 138 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
Use four enhancement iterations instead of one (ablation)MENet · OMRON, DUTS-TE, HKU-IS, PASCAL-S, ECSSD, SODMAE: 0.0617 -> 0.0450; 0.0496 -> 0.0281; 0.0504 -> 0.0234; 0.0868 -> 0.0535; 0.0673 -> 0.0307; 0.1449 -> 0.0874Improved
Use four enhancement iterations instead of three (ablation)MENet · OMRON, DUTS-TE, HKU-IS, PASCAL-S, ECSSD, SODMAE: 0.0504 -> 0.0450; 0.0389 -> 0.0281; 0.0383 -> 0.0234; 0.0685 -> 0.0535; 0.0503 -> 0.0307; 0.1147 -> 0.0874Improved
Add object-level loss to pixel- and region-level losses (ablation)MENet · OMRON, DUTS-TE, HKU-IS, PASCAL-S, ECSSDMAE: 0.0470 -> 0.0450; 0.0295 -> 0.0281; 0.0250 -> 0.0234; 0.0557 -> 0.0535; 0.0305 -> 0.0307Improved
Add object-level loss to pixel- and region-level losses (ablation)MENet · SODMAE: 0.0886 -> 0.0874Improved
Use four sub-regions instead of one in region-level loss (ablation)MENet · OMRON, DUTS-TE, HKU-IS, PASCAL-S, ECSSD, SODMAE: 0.0472 -> 0.0450; 0.0295 -> 0.0281; 0.0255 -> 0.0234; 0.0576 -> 0.0535; 0.0326 -> 0.0307; 0.0944 -> 0.0874Improved
Add gradient supervision to the multilevel hybrid loss (ablation)MENet · OMRON, DUTS-TE, HKU-IS, PASCAL-SMAE: 0.0470 -> 0.0450; 0.0295 -> 0.0281; 0.0250 -> 0.0234; 0.0557 -> 0.0535Improved
Add gradient supervision to the multilevel hybrid loss (ablation)MENet · ECSSDMAE: 0.0305 -> 0.0307Hurt
Add gradient supervision to the multilevel hybrid loss (ablation)MENet · SODMAE: 0.0886 -> 0.0874Improved
Combine region-level loss with pixel-level loss (ablation)MENet · OMRON, DUTS-TE, PASCAL-S, ECSSD, SODMAE: 0.0518 -> 0.0492; 0.0308 -> 0.0295; 0.0572 -> 0.0565; 0.0308 -> 0.0297; 0.0910 -> 0.0841Improved
Combine region-level loss with pixel-level loss (ablation)MENet · HKU-ISMAE: 0.0237 -> 0.0252Hurt

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