AI Paper Atlas

Research map / Detection and segmentation / CNN features

CompetitorFormer: Mitigating Query Conflicts for 3D Instance Segmentation via Competitive Strategy

Duanchu Wang, Junjie Yang, Haoran Gong, Jing Liu, Di Wang

CVPR 2026 · Detection and segmentation › CNN features · approach: Boundary-aware instance segmentation

Read the paper

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 Query Competition Layer (ablation)Relation3D-style baseline · ScanNetV2 valmAP: 62.2 -> 62.8Improved
add Relative Relationship Encoding (ablation)baseline · ScanNetV2 valmAP: 62.2 -> 62.7Improved
add Rank Cross-Attention (ablation)baseline · ScanNetV2 valmAP: 62.2 -> 62.6Improved
add QCL on top of RRE+RCA (ablation)CompetitorFormer · ScanNetV2 valmAP: 62.9 -> 63.4Improved
combine all three modules (ablation)baseline · ScanNetV2 valmAP: 62.2 -> 63.4Improved
increase RRE buckets L from 70 to 80 (ablation)RRE · ScanNetV2 valmAP: mAP 63.4 -> 63.1; mAP50 81.6 -> 81.8Mixed
use fewer RRE buckets with small width (L=50, v=0.010) (ablation)RRE · ScanNetV2 valmAP: 63.4 -> 62.6Hurt
add competition modelingbaseline · ScanNetV2 valconvergence: 62.0 mAP at epoch 300 vs baseline at epoch 450Improved
add competition modelingRelation3D · ScanNet200 valmAP: 31.6 -> 34.1Improved
add competition modelingDCD · ScanNet++V2 valmAP: 28.5 -> 34.1Improved

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