AI Paper Atlas

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

Domain Re-Modulation for Few-Shot Generative Domain Adaptation

Yi Wu, Ziqiang Li, Chaoyue Wang, Heliang Zheng, Shanshan Zhao, Bin Li, Dacheng Tao

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
remove the target mapping network (use frozen source mapping) (ablation)DoRM · 10-shot FFHQ-BabiesFID: 30.31 -> 86.52Hurt
remove the target mapping network (use frozen source mapping) (ablation)DoRM · 10-shot FFHQ-SunglassesFID: 17.31 -> 74.71Hurt
remove target affines in low-resolution layers (ablation)DoRM · 10-shot FFHQ-BabiesFID: 30.31 -> 93.28Hurt
remove target affines in low-resolution layers (ablation)DoRM · 10-shot FFHQ-SunglassesFID: 17.31 -> 92.42Hurt
remove target affines in high-resolution layers (ablation)DoRM · 10-shot FFHQ-BabiesFID: 30.31 -> 37.16Hurt
remove target affines in high-resolution layers (ablation)DoRM · 10-shot FFHQ-SunglassesFID: 17.31 -> 20.81Hurt
vary re-modulation weight alpha from 0.5 to 0.005 (ablation)DoRM · 10-shot FFHQ-BabyFID: 37.9 -> 30.3Improved
vary re-modulation weight alpha, sunglasses optimum at 0.2 (ablation)DoRM · 10-shot FFHQ-SunglassesFID: 18.7 -> 17.3Mixed
use one-layer instead of two-layer target domain classifier (ablation)DoRM · 10-shot FFHQ-BabyFID: 30.31 -> 37.15Hurt
use three-layer instead of two-layer target domain classifier (ablation)DoRM · 10-shot FFHQ-BabyFID: 30.31 -> 33.25Hurt
use source-initialized instead of random-initialized classifier (two layers) (ablation)DoRM · 10-shot FFHQ-SunglassesFID: 17.31 -> 20.40Hurt
replace full-generator fine-tuning with M&A modules for 10-domain generationDoRM vs CDC · multi-domain generationgenerator parameters: 240M -> 84MImproved

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