AI Paper Atlas

Research map / Label-efficient and robust learning / Domain adaptation and generalization

Learning Domain-Aware Detection Head with Prompt Tuning

Haochen Li, Rui Zhang, Hantao Yao, Xinkai Song, Yifan Hao, Yongwei Zhao, Ling Li, Yunji Chen

NeurIPS 2023 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Cross-domain few-shot with transformers · 6 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
learnable CoOp-style 16 domain-invariant tokens instead of hand-crafted prompt (ablation)DA-Pro · Cityscapes->Foggy CityscapesAP: 28.5 -> 28.9Improved
add domain-related textual description dl_d (ablation)DA-Pro · Cityscapes->Foggy CityscapesAP: 28.9 -> 29.2Improved
domain-specific tokens only (N=16) instead of domain-invariant (ablation)DA-Pro · Cityscapes->Foggy CityscapesAP: 29.2 -> 28.9No clear effect
split 16 tokens into 8 domain-invariant and 8 domain-specific (ablation)DA-Pro · Cityscapes->Foggy CityscapesAP: 29.2 -> 31.2Improved
add historical prompt ensemble (ablation)DA-Pro · Cityscapes->Foggy CityscapesAP: 31.2 -> 31.9Improved
vary tokens per type M=N from 4 to 16 (with ensemble) (ablation)DA-Pro · Cityscapes->Foggy CityscapesAP: 31.1 / 31.9 / 31.4 / 31.3Mixed
add loss terms progressively (L_s,t, L_s,st+L_t,st, L_ent) (ablation)DA-Pro · Cityscapes->Foggy CityscapesmAP: 54.4 -> 55.5Improved
add L_t,{s} loss term (ablation)DA-Pro · Cityscapes->Foggy CityscapesmAP: 55.5 -> 53.4Hurt
add historical prompt ensemble across loss forms (ablation)DA-Pro · Cityscapes->Foggy CityscapesmAP: 55.5 -> 55.9Improved
replace hard pseudo labels with probability (soft) supervision (ablation)DA-Pro · Cityscapes->Foggy CityscapesmAP: 55.9 -> 54.3Hurt
replace hard pseudo labels with probability (soft) supervision (ablation)DA-Pro · SIM10K->CityscapesmAP: 62.9 -> 62.1Hurt
remove domain-related textual tokens dl_d (ablation)DA-Pro · Cityscapes->Foggy CityscapesmAP: 55.9 -> 54.3Hurt

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