AI Paper Atlas

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

Recognizing Part Attributes With Insufficient Data

Xiangyun Zhao, Yi Yang, Feng Zhou, Xiao Tan, Yuchen Yuan, Yingze Bao, Ying Wu

ICCV 2019 · Label-efficient and robust learning › Zero-/few-shot and meta-learning · approach: Generative feature synthesis for zero-shot · 17 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 part-localization sharing (ablation)CSN, 200 train imgs · CUB bill-greyAP: 38.5% -> 42.2%Improved
add pattern sharing on top of part sharing (ablation)CSN, 200 train imgs · CUB bill-greyAP: 42.2% -> 46.8%Improved
add pattern sharing on top of part sharing (ablation)CSN, 500 train imgs · CUB bill-blackAP: 79.4% -> 77.0%Hurt
add pattern sharing on top of part sharing (ablation)CSN, 200 train imgs · CUB bill-blackAP: 78.6% -> 76.0%Hurt
add part sharing (ablation)CSN, 500 train imgs · CUB belly-solidAP: 83.6% -> 85.9%Improved
add part sharing (ablation)CSN, 200 train imgs · CUB belly-solidAP: 79.2% -> 84.0%Improved
increase number of sharing part attributes from 0 to 24 (ablation)CSN · CUB bill-orangeAP: 38.6% -> 42.7%Improved
add 9 sharing pattern attributes on top of 24 part attributes (ablation)CSN · CUB bill-orangeAP: 42.7% -> 43.0%Improved
add concept sharing (CSN) instead of independent branchesBaseline · SurveilA attribute 10 (umbrella)AP: 0.5% -> 30.3%Improved
add concept sharing (CSN) instead of independent branchesBaseline · SurveilAmAP: 30.3% -> 51.2%Improved
add concept sharing localization to CSN w/o share (ablation)CSN w/o share · CelebAattribute accuracy: 91.7% -> 91.8%Improved
initialise soft part weights to all ones instead of one-hot prior (ablation)CSN-soft · CUB-200-2011 (32 attributes)AP: 65.1% -> 63.8%Hurt

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