AI Paper Atlas

Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning

Attribute Prototype Network for Any-Shot Learning

Wenjia Xu, Yongqin Xian, Jiuniu Wang, Bernt Schiele, Zeynep Akata

IJCV 2022 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Generative feature synthesis for zero-shot · 47 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 attribute regression loss L_Reg with ProtoMod (ablation)BaseMod · CUB ZSLtop-1 accuracy: 70.0 -> 71.5Improved
add attribute decorrelation loss L_AD (ablation)BaseMod + L_Reg · AWA2 ZSLtop-1 accuracy: 66.3 -> 67.7Improved
add attribute decorrelation loss L_AD (ablation)BaseMod + L_Reg · CUB part localizationPCP mean: 35.2 -> 49.0Improved
add compactness regularizer L_CPT (ablation)BaseMod + L_Reg + L_AD · ZSL CUB, AWA2, SUNtop-1 accuracy: 71.8 -> 72.0No clear effect
add compactness regularizer L_CPT (ablation)BaseMod + L_Reg + L_AD · CUB part localizationPCP mean: 49.0 -> 52.8Improved
add ZoomInMod (ablation)APN without ZoomInMod · CUB ZSLtop-1 accuracy: 72.0 -> 75.0Improved
add ZoomInMod (ablation)APN without ZoomInMod · SUN ZSLtop-1 accuracy: 61.6 -> 61.5No clear effect
add ZoomInMod (ablation)APN · compute costFLOPs (GMac): 7.9 -> 15.7Hurt
use APN features instead of finetuned features for ABPgenerative ZSL model · AWA2 ZSLtop-1 accuracy: 68.5 -> 73.8Improved
use APN features instead of finetuned features for CVCgenerative ZSL model · AWA2 ZSLtop-1 accuracy: 64.6 -> 71.2Improved
use APN features for ABPgenerative ZSL model · SUN ZSLtop-1 accuracy: 62.6 -> 63.1Improved
use APN features for CVCgenerative GZSL model · AWA2 GZSL unseen / seenaccuracy: gains 5.8% on unseen while sacrificing only 2.1% on seenImproved

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