AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

SCoA: Revisiting Domain Generalized Object Detection with Style-Conditioned Adaptation

Han Jiang, Wenfei Yang, Tianzhu Zhang, Yongdong Zhang

ICML 2026 · Detection and segmentation › Open-set and few-shot detection · approach: Domain-adaptive object detection

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
add Spectral Style Modeling to baseline (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 55.3 -> 55.7Improved
add Mixture-of-Tokens Adaptation (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 55.7 -> 58.7Improved
add Style-Conditioned Query Refinement (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 58.7 -> 59.4Improved
use adversarial style sampling instead of random (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 58.9 -> 59.4Improved
add style memory (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 58.7 -> 59.4Improved
add shared matrix A in tokens (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 58.5 -> 59.4Improved
use low-rank parameterization of tokens (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 58.9 -> 59.4Improved
use style-conditioned router rather than random (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 56.4 -> 59.4Improved
add adaptive multi-layer token fusion over final-layer tokens only (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 59.0 -> 59.4Improved
condition layer fusion weights on style vector (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 59.1 -> 59.4Improved
refine DETR queries via cross-attention instead of replacing with learnable tokens (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 54.7 -> 59.4Improved
use mixture-of-tokens instead of token-level modulation (ablation)SCoA · Urban Scene ablation (Avg of DF/DR/NR/NC)mAP@0.5 (Avg): 58.1 -> 59.4Improved

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