AI Paper Atlas

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

Task-Independent Knowledge Makes for Transferable Representations for Generalized Zero-Shot Learning

Chaoqun Wang, Xuejin Chen, Shaobo Min, Xiaoyan Sun, Houqiang Li

AAAI 2021 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Generative feature synthesis for zero-shot · 22 citations (OpenAlex)

Read the paper

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 VCM instance discrimination to Basic-ZSL (ablation)Basic-ZSL · CUBH: 63.5 -> 68.5Improved
increase the instance-discrimination loss weight from 0 to 0.1 (ablation)Basic-ZSL with VCM · CUB and aPYHImproved
increase the instance-discrimination loss weight from 0.1 to 1.0 (ablation)Basic-ZSL with VCM · CUB and aPYHHurt
add attribute masking (ablation)Basic-ZSL with VCM · CUBH: 68.5 -> 69.5Improved
add cross-modal triplet loss (ablation)VCM with attribute masking · CUBH: 69.5 -> 69.9; 0.4% improvementImproved
add masked attribute prediction loss (ablation)VCM with attribute masking and cross-modal triplet · CUBH: 69.9 -> 70.4Improved
add grayscale augmentation (ablation)VCM with cropping and horizontal flipping · CUBH: 66.0 -> 65.5Hurt
add color jittering (ablation)VCM with cropping and horizontal flipping · CUBH: 66.0 -> 63.8; 66.0 -> 65.3; 66.0 -> 65.2Hurt
add Gaussian blurring (ablation)VCM with cropping and horizontal flipping · CUBH: 66.0 -> 67.1Improved
add rotation; larger tested angles perform worse than the 30-degree setting (ablation)VCM with cropping and horizontal flipping · CUBH: 66.0 -> 64.4; 66.0 -> 64.7; 66.0 -> 66.5Mixed
add jigsaw swapping; larger tested patch counts perform worse than 3 (ablation)VCM with cropping and horizontal flipping · CUBH: 66.0 -> 64.2; 66.0 -> 65.8; 66.0 -> 66.7Mixed
combine positive low-level augmentations (ablation)Basic-ZSL with VCM · CUBH: 65.3 -> 68.5; 3.2%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