AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Fine-Grained Retrieval Prompt Tuning

Shijie Wang, Jianlong Chang, Zhihui Wang, Haojie Li, Wanli Ouyang, Qi Tian

AAAI 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 25 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
Use DPP with a frozen pre-trained model (ablation)Pre-training (PT) · CUB-200-2011Recall@1: improves the Recall@1 accuracy by 10.8%Improved
Use DPP with a frozen pre-trained model (ablation)Pre-training (PT) · Stanford CarsRecall@1: 15.8%Improved
Add CAH to the frozen pre-trained model (ablation)Pre-training (PT) · CUB-200-2011Recall@1: 16.7%Improved
Add CAH to the frozen pre-trained model (ablation)Pre-training (PT) · Stanford CarsRecall@1: 18.0%Improved
Combine DPP and CAH (ablation)Frozen pre-trained model · CUB-200-2011Recall@1: 29.4%Improved
Combine DPP and CAH (ablation)Frozen pre-trained model · Stanford CarsRecall@1: 32.2%Improved
Add Instance Normalization within CAH (ablation)DPP + PT + CAH without IN · CUB-200-2011Recall@1: 72.2% -> 74.3%Improved
Add Instance Normalization within CAH (ablation)DPP + PT + CAH without IN · Stanford CarsRecall@1: 89.6% -> 91.1%Improved
Use FRPT instead of full fine-tuningResNet-50 retrieval model · CUB-200-2011Recall@1: 4.6%Improved
Use FRPT instead of full fine-tuningResNet-50 retrieval model · Stanford CarsRecall@1: 6.7%Improved
Use discriminative perturbation prompting instead of fixed CAM or bounding-box cropping (ablation)CUB-200-2011 retrieval · CUB-200-2011Recall@1: 63.7% -> 74.3%; 67.6% -> 74.3%Improved
Use FRPT instead of fine-tuningResNet-50 in the few-shot setting · CUB-200-2011 10-shotRecall@1: 63.1% -> 66.6%Improved

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