AI Paper Atlas

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

Semantic-guided Robustness Tuning for Few-Shot Transfer Across Extreme Domain Shift

kangyu xiao, Zilei Wang, junjie li

ECCV 2024 · Label-efficient and robust learning › Domain adaptation and generalization · approach: Cross-domain few-shot with transformers · 1 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
add SRT fine-tuning to CLIPCLIP ViT-B · BSCD-FSL 1-shot avg cls-Iaccuracy: 50.93 -> 57.12Improved
traditional fine-tuning of the last block on DINODINO ViT-B · BSCD-FSL 1-shot avg cls-Iaccuracy: -14.07Hurt
full fine-tuning CLIP with Trad-FTCLIP ViT-B · BSCD-FSL 1-shot avg cls-Iaccuracy: -5.31Hurt
add L_RI loss (ablation)SRT on CLIP · EuroSAT 1-shotaccuracy: 68.28 -> 73.72Improved
add L_CLIP on top of L_RI (ablation)SRT on CLIP · EuroSAT 1-shotaccuracy: 73.72 -> 81.03Improved
add semantic injection via MMIT-Mixup versus none (ablation)SRT on CLIP · EuroSAT 1-shot cls-Taccuracy: 65.28 -> 74.95Improved
replace MMIT-Mixup with cross-attention injection on DINO (ablation)SRT on DINO · ISIC 1-shot cls-Iaccuracy: 35.64 -> 32.98Hurt
cross-attention injection versus none on CLIP (ablation)SRT on CLIP · EuroSAT 1-shot cls-Iaccuracy: 72.45 -> 73.29Improved
MMIT-Mixup versus cross-attention in extra resources (ablation)SRT · FLOPs and paramsextra parameters: 2.16 M -> 0Improved
apply SRT to lightweight DINO ViT-SDINO ViT-S · BSCD-FSL 1-shot avg cls-Iaccuracy: 51.98 -> 55.33Improved

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