AI Paper Atlas

Research map / Detection and segmentation / CNN features

Symmetry-Aware Transformer-Based Mirror Detection

AAAI 2023 · Detection and segmentation › CNN features · approach: Context for hard detection conditions · 33 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (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 SATNet with loose-symmetry modeling instead of the strongest reported baselineMirror detection · MSDIoU: 81.54 -> 85.41Helped
Use SATNet without depth instead of the strongest reported depth baselineMirror detection · RGBD-MirrorIoU: 77.77 -> 78.42Helped
Add separately supervised dual-path features without suitable fusion or matching (ablation)Swin-S mirror detector · MSDIoU: 80.46 -> 79.59Hurt
Add SAAM at scale 3 (ablation)Dual-Path Swin-S · MSDFβ: 0.903 -> 0.918Helped
Apply SAAM at the lower scale as well as scale 3 (ablation)Dual-Path + SAAM · MSDFβ: 0.918 -> 0.903Hurt
Replace UperNet with CFDM (ablation)Dual-Path Swin-S · MSDIoU: 79.59 -> 81.98Helped
Combine scale-3 SAAM with CFDM (ablation)Dual-Path Swin-S · MSDIoU: 81.98 -> 82.96Helped
Add SAAM at both high scales and use CFDM (ablation)Dual-Path Swin-S · MSDIoU: 80.03 -> 85.41Helped

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