AI Paper Atlas

Research map / Detection and segmentation / CNN features

MT-ORL: Multi-Task Occlusion Relationship Learning

Panhe Feng, Qi She, Lei Zhu, Jiaxin Li, Lin Zhang, Zijian Feng, Changhu Wang, Chunpeng Li, Xuejing Kang, Anlong Ming

ICCV 2021 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 9 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
share top-2 stages rather than top-1 stage (ablation)OPNet · PIODboundary AP: 80.8 -> 83.1Improved
share top-2 stages rather than top-3 stages (ablation)OPNet · PIODboundary AP: 80.6 -> 83.1Improved
share top-2 stages rather than top-4 stages (ablation)OPNet · PIODboundary AP: 79.7 -> 83.1Improved
add orthogonal occlusion representation (OOR) in place of dbr orientation loss (ablation)OPNet · PIODorientation AP: 75.9 -> 79.4Improved
add OOR in place of dbr orientation loss (ablation)OPNet · BSDS ownershiporientation AP: 50.5 -> 60.1Improved
add OOR in place of dor orientation loss (ablation)OPNet · PIODorientation AP: 72.2 -> 79.4Improved
add OOR in place of dbr orientation loss (ablation)OPNet · BSDS ownershipboundary AP: 62.4 -> 66.8Improved
use five-scale supervision rather than one side map (ablation)OPNet boundary path · PIODboundary AP: 77.4 -> 83.1Improved
use five-scale supervision rather than four (ablation)OPNet boundary path · PIODboundary AP: 81.1 -> 83.1Improved
use two-scale supervision rather than three (ablation)OPNet boundary path · PIODboundary AP: 82.6 -> 82.0No clear effect
shared-high, separated-low OPNet architecture with dbr loss vs OFNetOPNet + dbr · PIODboundary AP: 77.0 -> 80.6Improved
train joint branches instead of boundary branch onlyOPNet · PIODboundary AP: 79.5 -> 83.1Improved

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